Research Handbook On Intellectual Property And Artificial Intelligence Ryan Abbott

Research Handbook On Intellectual Property And
Artificial Intelligence Ryan Abbott download
https://ebookbell.com/product/research-handbook-on-intellectual-
property-and-artificial-intelligence-ryan-abbott-49840310
Explore and download more ebooks at ebookbell.com

Here are some recommended products that we believe you will be
interested in. You can click the link to download.
Research Handbook On Intellectual Property And Cultural Heritage Irini
Stamatoudi Editor
property-and-cultural-heritage-irini-stamatoudi-editor-50288674
Research Handbook On Intellectual Property And Competition Law Josef
Drexl
property-and-competition-law-josef-drexl-2480068
Research Handbook On Intellectual Property And Investment Law Research
Handbooks In Intellectual Property Series Christophe Geiger Editor
property-and-investment-law-research-handbooks-in-intellectual-
property-series-christophe-geiger-editor-38503602
Research Handbook On Intellectual Property And Technology Transfer
Jacob H Rooksby
property-and-technology-transfer-jacob-h-rooksby-43476282

Research Handbook On Intellectual Property And Employment Law N Bruun
property-and-employment-law-n-bruun-43476292
Research Handbook On Intellectual Property And Digital Technologies
Tanya Aplin
property-and-digital-technologies-tanya-aplin-43476296
Research Handbook On Intellectual Property And Creative Industries
Abbe E L Brown
property-and-creative-industries-abbe-e-l-brown-43476304
Research Handbook On The World Intellectual Property Organization The
First 50 Years And Beyond Sam Ricketson
https://ebookbell.com/product/research-handbook-on-the-world-
intellectual-property-organization-the-first-50-years-and-beyond-sam-
ricketson-43476068
Research Handbook On The Interpretation And Enforcement Of
Intellectual Property Under Wto Rules Intellectual Property In The Wto
Research Handbooks Carlos M Correa
https://ebookbell.com/product/research-handbook-on-the-interpretation-
and-enforcement-of-intellectual-property-under-wto-rules-intellectual-
property-in-the-wto-research-handbooks-carlos-m-correa-2434626

RESEARCH HANDBOOK ON INTELLECTUAL
PROPERTY AND ARTIFICIAL INTELLIGENCE

RESEARCH HANDBOOKS IN INTELLECTUAL PROPERTY
Series Editor: Jeremy Phillips, Intellectual Property Consultant, Olswang, Research Director,
Intellectual Property Institute and co-founder, IPKat weblog
Under the general editorship and direction of Jeremy Phillips comes this important new
Research Handbook series of high quality, original reference works that cover the broad
pillars of intellectual property law: trademark law, patent law and copyright law – as well
as less developed areas, such as geographical indications, and the increasing intersection of
intellectual property with other fields. Taking an international and comparative approach,
these Research Handbooks, each edited by leading scholars in the respective field, will com-
prise specially commissioned contributions from a select cast of authors, bringing together
renowned figures with up-and-coming younger authors. Each will offer a wide-ranging
examination of current issues in intellectual property that is unrivalled in its blend of critical,
innovative thinking and substantive analysis, and in its synthesis of contemporary research.
Each Research Handbook will stand alone as an invaluable source of reference for all
scholars of intellectual property, as well as for practising lawyers who wish to engage with the
discussion of ideas within the field. Whether used as an information resource on key topics, or
as a platform for advanced study, these Research Handbooks will become definitive scholarly
reference works in intellectual property law.
Titles in the series include:
Research Handbook on Intellectual Property and Digital Technologies
Edited by Tanya Aplin
Research Handbook on Intellectual Property and Technology Transfer
Edited by Jacob H. Rooksby
Research Handbook on Intellectual Property and Investment Law
Edited by Christophe Geiger
Research Handbook on the World Intellectual Property Organization
The First 50 Years and Beyond
Edited by Sam Ricketson
Research Handbook on Trademark Law Reform
Edited by Graeme B. Dinwoodie and Mark D. Janis
Research Handbook on Design Law
Edited by Henning Hartwig
Research Handbook on Intellectual Property and Employment Law
Edited by Niklas Bruun and Marja-Leena Mansala
Research Handbook on Intellectual Property and Cultural Heritage
Edited by Irini Stamatoudi
Research Handbook on Intellectual Property and Artificial Intelligence
Edited by Ryan Abbott

Cheltenham, UK • Northampton, MA, USA
RESEARCH HANDBOOKS IN INTELLECTUAL PROPERTY
Research Handbook on Intellectual
Property and Artificial Intelligence
Edited by
Ryan Abbott
Professor of Law and Health Sciences, School of Law, University of Surrey,
UK and Adjunct Assistant Professor of Medicine, David Geffen School of
Medicine, University of California Los Angeles, USA

© The Editor and Contributors Severally 2022
Cover image: Conceived by DABUS, complements of Dr. Stephen Thaler
All rights reserved. No part of this publication may be reproduced, stored in a retrieval system
or transmitted in any form or by any means, electronic, mechanical or photocopying, recording,
or otherwise without the prior permission of the publisher.
Published by
Edward Elgar Publishing Limited
The Lypiatts
15 Lansdown Road
Cheltenham
Glos GL50 2JA
UK
Edward Elgar Publishing, Inc.
William Pratt House
9 Dewey Court
Northampton
Massachusetts 01060
USA
A catalogue record for this book
is available from the British Library
Library of Congress Control Number: 2022946033
This book is available electronically in the
Law subject collection
http://dx.doi.org/10.4337/9781800881907
ISBN 978 1 80088 189 1 (cased)
ISBN 978 1 80088 190 7 (eBook)

v
Contents
List of contributorsvii
PART I MULTI-SUBJECT
1 Intellectual property and artificial intelligence: an introduction 2
Ryan Abbott
2 The human cause 22
Daniel J. Gervais
3 Considering intellectual property law for embodied forms of artificial
intelligence40
Woodrow Barfield, Argyro Karanasiou and Karni Chagnal-Feferkorn
4 AI replication of musical styles points the way to an exclusive rights regime 65
Sean M. O’Connor
5 The elusive intellectual property protection of trained machine learning
models: a European perspective 84
Jean-Marc Deltorn
6 An abject failure of intelligence: intellectual property and artificial intelligence 113
Michael D. Pendleton
PART II COPYRIGHT AND RELATED RIGHTS
7 The AI–copyright challenge: tech-neutrality, authorship, and the public interest 134
Carys J. Craig
8 Four theories in search of an A(I)uthor 156
Giancarlo Frosio
9 Copyright law should stay true to itself in the age of artificial intelligence 179
Alice Lee and Phoebe Woo
10 The protection of AI-generated pictures (photograph and painting)
under copyright law 198
Yaniv Benhamou and Ana Andrijevic
11 Performers’ rights and artificial intelligence 218
Richard Arnold
12 AIn’t it just software? 225
Shubha Ghosh

vi Research handbook on intellectual property and artificial intelligence
13 Can artificial intelligence infringe copyright? Some reflections 245
Enrico Bonadio, Plamen Dinev and Luke McDonagh
PART III TRADE MARKS AND DESIGNS
14 Computational trademark infringement and adjudication 259
Daryl Lim
15 Online shopping with artificial intelligence: what role for trade marks? 290
Anke Moerland and Christie Kafrouni
16 Trademark law, AI-driven behavioral advertising, and the Digital
Services Act: toward source and parameter transparency for consumers,
brand owners, and competitors 309
Martin Senftleben
17 A quotidian revolution: artificial intelligence and trade mark law 325
Dev S. Gangjee
18 The impact of AI on designs law 346
Trevor Cook
PART IV PATENTS AND TRADE SECRETS
19 Legal fictions and the corporation as an inventive artificial intelligence 356
Dennis Crouch
20 Economic reasons to recognise AI inventors 376
Benjamin Mitra-Kahn
21 Reverse engineering (by) artificial intelligence 391
Shawn Bayern
22 Trade secrets versus the AI explainability principle 405
Rita Matulionyte and Tatiana Aranovich
23 The inventive step requirement and the rise of the AI machines 423
Noam Shemtov and Garry A. Gabison
24 Trade secrecy, factual secrecy and the hype surrounding AI 443
Sharon K. Sandeen and Tanya Aplin
Index461

vii
Contributors
Ryan Abbott is Professor of Law and Health Sciences at University of Surrey School of
Law and Adjunct Assistant Professor of Medicine at the David Geffen School of Medicine
at UCLA. Ryan is the author of The Reasonable Robot: Artificial Intelligence and the Law,
published in 2020 by Cambridge University Press. He has written widely on issues associated
with law and technology, health law and intellectual property in leading legal, medical and
scientific books and journals. Ryan’s research has been featured prominently in the media,
including the New York Times, Wall Street Journal and Financial Times. He is a licensed phy-
sician, patent attorney and acupuncturist in the United States, a solicitor advocate in England
and Wales, and board-certified by the American Board of Legal Medicine (ABLM). Managing
Intellectual Property magazine named him as one of the 50 most influential people in intellec-
tual property in 2019 and again in 2021.
Ana Andrijevic is a PhD candidate and a research and teaching assistant at the University of
Geneva, Faculty of Law/Digital Law Center. Her doctoral dissertation on the impact of AI on
copyright law is being written under the supervision of Prof. Jacques de Werra. Her areas of
research focus on technology law (including AI and children’s rights in the digital age) and
intellectual property law.
Tanya Aplin is Professor of Intellectual Property Law at the Dickson Poon School of Law,
King’s College London. She has published widely on copyright, confidentiality and trade
secrets law. Some of her leading co-authored publications include Gurry on Breach of
Confidence: The Protection of Confidential Information (OUP, 2012), Intellectual Property:
Patents, Copyright, Trade Marks and Allied Rights 9th ed (Sweet Maxwell, 2019), Global
Mandatory Fair Use: The Nature and Scope of the Right to Quote Copyright Works (CUP,
2020) and a Research Handbook on IP and Digital Technologies (Edward Elgar Publishing,
2020).
Tatiana Aranovich is Assistant to the Director at the Healthcare Insurance Authority. She
holds a Master’s degree in Law (MPhil) and an undergraduate degree in Economics, both
from Universidade Federal do Rio Grande do Sul (UFRGS), and a law degree from Pontificia
Universidade Católica – PUC. She served as an internee at the Antitrust Authority (OFT,
currently CMA) in the UK.
Richard Arnold read Chemistry at the University of Oxford before being called to the Bar of
England and Wales in 1985. He specialized in intellectual property law and became a QC in
2000. He was a Judge of the High Court, Chancery Division from October 2008 to September
2019 and Judge in Charge of the Patents Court from April 2013 to September 2019. He has
been an External Member of the Enlarged Board of Appeal of the European Patent Office
since March 2016 and a Judge of the Court of Appeal since October 2019. He is the author of
Performers’ Rights (6th ed, Sweet Maxwell, 2021) and the editor of the Halsbury’s Laws
of England title Trade Marks and Trade Names (5th ed, Butterworths, 2014). He was editor

viii Research handbook on intellectual property and artificial intelligence
of Entertainment and Media Law Reports from 1993 to 2004 inclusive and has published
numerous articles in legal journals.
Woodrow Barfield, Whitaker Institute and Visiting Professor of Law, University of Turin,
holds a PhD in engineering and a JD and LLM in intellectual property law and policy. He
received the Presidential Young Investigator award from the National Science and is the author
of Cyber Humans: Our Future with Machines, co-author (with Ugo Pagallo) of Advanced
Introduction to Law and Artificial Intelligence, co-editor (with Ugo Pagallo) of Research
Handbook on Law and AI, and co-editor (with Marc Jonathan Blitz) of Research Handbook
of Law on Virtual and Augmented Reality. He is also the editor of the Cambridge Handbook
of Law of Algorithms and editor of Fundamentals of Wearable Computers and Augmented
Reality. His current interests are the law as related to human enhancement technology, arti-
ficial intelligence and algorithms, and, from an engineering perspective, the computational
infrastructure of the human body.
Shawn Bayern is Larry Joyce Beltz Professor and Associate Dean for Academic Affairs at
the Florida State University College of Law. His teaching and research focus on common-law
issues, primarily in contracts, torts and organizational law. Professor Bayern has visited at
Berkeley Law, Duke Law School and the Northwestern Pritzker School of Law. He is the
author, most recently, of Autonomous Organizations (2021, Cambridge University Press).
Bayern is an elected member of the American Law Institute and serves as advisor to part of
the Restatement (Third) of Torts. He has an extensive background in computer science and,
before his legal academic career, designed a widely used identity authentication protocol and
served on the specification expert groups for several computer-programming languages. He
has a JD from Berkeley Law and a BS in computer science from Yale University.
Yaniv Benhamou is Associate Professor of Digital Law at the University of Geneva,
Faculty of Law/Digital Law Center, and attorney-at-law at a Swiss-based law firm. Professor
Benhamou specializes in data protection, intellectual property, and art law, was appointed as
an expert by the World Intellectual Property Organization (WIPO) for copyright and museums
and regularly teaches copyright law to museums in the framework of the legal training of
AMS (Association of Swiss Museums). In addition to these legal activities, he participates
in associative and cultural activities in the fields of art and music. In particular, he founded
lab-of-arts and Artists Rights, free legal consultations for Swiss artists (that is, Swiss lawyers
volunteering for the arts). He is Legal Affairs Committee Member of the International Council
of Museums (ICOM).
Enrico Bonadio is Reader in Intellectual Property (IP) law at City, University of London. His
research agenda has recently focused on – inter alia – the intersection between IP and technol-
ogy, including the impact of artificial intelligence (AI) and robotics innovation on copyright
and patent laws. Enrico is member of the Centre for Creativity enabled by AI, funded by
UKRI’s Research England and City, University of London. He has been part of an EU-funded
group of researchers to assess the area of interactive robots in society. Enrico is also Deputy
Editor in Chief of the European Journal of Risk Regulation.
Karni Chagnal-Feferkorn is a postdoctoral fellow in AI and Regulation at the University
of Ottawa. Her research examines different aspects of the intersection between AI and the
law, including legal liability for AI-induced damages, governance by AI and the educational

Contributors ix
challenge of teaching lawyers and data scientists to work jointly in order to design responsible
AI systems. Karni pursued her PhD at the University of Haifa (Israel) and holds an LLM in
Law, Science and Technology from Stanford University. She is a licensed attorney in Israel,
California and New York. In addition to her academic research, Karni is a founding partner
of a consultancy firm that specializes in comparative research pertaining to law and regula-
tion, and conducts research for governments, law firms and companies on various regulatory
matters, including technology in general and AI specifically. She also advises an Israeli
start-up focused on automating the drafting of complex legal documents.
Trevor Cook is a partner at WilmerHale, where he focuses his practice on transnational
intellectual property litigation matters and is also active in the area of life sciences. Mr Cook
has more than 40 years of experience in global patent litigation, particularly in Europe and
Asia. He has acted in many of the leading patent infringement cases that have come before
the English courts, many of which have concerned pharmaceuticals and biotechnology, and
also in some of the leading cases regarding the protection of regulatory data. Mr Cook joined
WilmerHale from Bird Bird LLP in London, where he was a partner in the Intellectual
Property Group and co-head of the International Life Sciences Sector Group.
Carys J. Craig is Associate Professor at Osgoode Hall Law School, York University, in
Toronto, Canada. She is Academic Director of the Professional LLM Program in Intellectual
Property Law, Editor-in-Chief of the Osgoode Hall Law Journal and recently served as
Associate Dean (Research Institutional Relations). Dr Craig teaches and publishes in the
areas of copyright, trade marks, law and technology and critical legal and feminist theory.
She is the author of Copyright, Communication Culture: Towards a Relational Theory of
Copyright Law (Edward Elgar Publishing, 2011). Dr Craig’s scholarship has been cited with
approval in several landmark copyright rulings by the Supreme Court of Canada. She holds
an LLB (First Class Honours) from the University of Edinburgh, an LLM from Queen’s
University in Kingston, Ontario and an SJD from the University of Toronto.
Dennis Crouch is Associate Professor of Law at the University of Missouri School of Law.
Prior to joining the MU Law Faculty, he was a patent attorney at McDonnell Boehnen Hulbert
Berghoff LLP in Chicago, Illinois, and taught at Boston University Law School. He has
worked on cases involving various technologies including computer memory and hardware,
circuit design, software, networking, mobile and internet telephony, automotive technologies,
lens design, bearings, HVAC systems and business methods. He is also the editor of the
popular patent law weblog Patently-O.
Jean-Marc Deltorn is Assistant Professor and Senior Researcher at the International Centre
for the Study of Intellectual Property (CEIPI), University of Strasbourg. At CEIPI, Jean-Marc
studies the interplay between IP norms and emerging digital technologies. He is co-founder
and director of CEIPI’s AI IP University Diploma and Adjunct Director of CEIPI’s Research
Laboratory. Prior to joining CEIPI, Jean-Marc spent more than 15 years at the European
Patent Office (EPO), where he chaired the prosecution of AI applications. Jean-Marc is
currently a member of the Impact of Technology Expert Group at the European Observatory
on Infringements of Intellectual Property Rights (EUIPO). Jean-Marc holds a PhD in Physics
from Paris University and a PhD in Law from Strasbourg University.
Plamen Dinev is Lecturer in Law at Goldsmiths, University of London. He holds a PhD

x Research handbook on intellectual property and artificial intelligence
from The City Law School and an LLM from Leiden University. His research focuses on the
relationship between IP law and new technology, including AI, 3D printing and digitization
more broadly. He has published widely in leading IP journals, including IPQ and EIPR, and
has presented his research at universities including Bocconi, Cardiff and Keio. In 2018 Plamen
received a Modern Law Review Scholarship, which allowed him to undertake fieldwork in
New York as part of an empirical project on 3D printing and the law. In 2019 he was invited to
the European Commission’s final report on the IP implications of 3D printing.
Giancarlo Frosio is Professor of Law Technology at the School of Law of Queen’s
University Belfast. He is also a Non-resident Fellow at Stanford Law School CIS, Stanford
University and Faculty Associate at Nexa Center, Polytechnic and University of Turin.
Garry A. Gabison is Senior Lecturer at the Centre for Commercial Law Studies, Queen
Mary University of London. Prior to joining the Centre, Garry taught economics and public
policy at the School of Public Policy and the School of Economics at the Georgia Institute
of Technology. His teaching focused on innovation policy. Garry also worked at the Joint
Research Centre (JRC), a Directorate General of the European Commission. At the JRC, Garry
investigated issues related to the Information and Communications Technology sector. His
work focused on innovation incentives and financing. Garry holds a JD from the University
of Virginia and a PhD in Economics from Yale University. He has published in journals in
Europe and in the United States on such issues as copyright, patents and AI.
Dev S. Gangjee is Professor of Intellectual Property Law at the University of Oxford and
Director of the Oxford IP Research Centre (OIPRC). He has a longstanding interest in trade
marks, geographical indications and copyright. Recent research has looked at the interface
between trade marks and innovation, as well as the influence of machine learning technologies
on trade mark registration.
Daniel J. Gervais PhD is the Milton R. Underwood Chair in Law at Vanderbilt University
Law School, where he serves as Director of the Vanderbilt Intellectual Property Program and
Co-director of the LLM Program. Prior to joining Vanderbilt, he was the Acting Dean at the
Faculty of Law of the University of Ottawa (Common Law Section). Before he joined the
Academy, Professor Gervais was Legal Officer at the GATT (now WTO), Head of Section at
WIPO and Vice-president of Copyright Clearance Center, Inc. (CCC). In 2012, he was elected
to the Academy of Europe. He is a member of the American Law Institute, where he serves
as Associate Reporter on the Restatement of the Law, Copyright Project. In 2022, he is the
immediate Past President of the International Association for the Advancement of Teaching
and Research in Intellectual Property (ATRIP).
Shubha Ghosh PhD, JD, is Crandall Melvin Professor of Law at Syracuse University College
of Law and Director of the Technology Commercialization Law Program and the Syracuse
Intellectual Property Law Institute.
Christie Kafrouni is a legal affairs and IP advisor for IMPS The Smurfs, Belgium. She holds
a Masters degree in Law from Université Catholique de Louvain (Belgium) and an Advanced
Masters in Intellectual Property and Knowledge Management from Maastricht University.
During her Masters studies at Maastricht University, she became acquainted with the subject
of online shopping and the challenges it causes for trade mark law; this resulted in a Masters
thesis supervised by Dr Anke Moerland.

Contributors xi
Argyro Karanasiou is Assistant Professor in Law and Innovation at the University of
Birmingham, the founding director of LETS Lab (Law, Emerging Tech Science) and
a research affiliate at the BHRE, University of Greenwich, London, where she previously
worked as Associate Professor in Information Technology Law. Her work contributes to the
growing body of transdisciplinary scholarship on law and emerging technologies and has
earned her visiting research affiliations with Yale Law School (ISP Alumna), NYU Law (ILI
Alumna), Harvard Law (affiliate Faculty staff CopyX), and Complutense Madrid (ITC). She is
actively involved in several technology policy and related initiatives in the UK and worldwide,
including those concerned with the regulation of AI, one of her key research interests.
Alice Lee, LLB (HKU), BCL (Oxford), is Associate Professor of Law at the University of
Hong Kong and Senior Fellow of AdvanceHE (SFHEA). She has received three HKU teach-
ing awards, including the University Distinguished Teaching Award, and chaired the HKU
Teaching Exchange Fellowship Sub-Group. She specializes in intellectual property and real
property and is the author of textbooks and practitioners’ texts in both areas. She co-founded
Creative Commons Hong Kong in 2008, initiated an IP Ambassador Programme with the
Intellectual Property Department of the HKSAR Government in 2016 and launched an online
animation series, ‘The Copyright Classroom’, in 2019 to support knowledge exchange and
public education. She has served on consultative committees and statutory bodies, including
the Advisory Committee on the Review of the Patent System, which advised the government
on the enactment of the Patents (Amendment) Ordinance 2016 for an original grant patent
system in Hong Kong.
Daryl Lim is Professor of Law and Director of the Center for Intellectual Property (IP),
Information Privacy Law at the University of Illinois Chicago School of Law. The IP Center
is a founding IP institution in the United States and is consistently ranked as offering one of the
premier IP programmes in the country. Professor Lim is an award-winning author, observer
and commentator on global trends in IP and competition policy and how they influence and
are influenced by law, technology, economics and politics. He regularly engages senior gov-
ernment officials, corporate leaders, civil society organizations, and law firms at national and
international conferences.
Rita Matulionyte, PhD, LLM is Senior Lecturer at Macquarie Law School, Macquarie
University (Australia) and Associate Senior Research Fellow at the Law Institute of Lithuania.
Her research is in the area of intellectual property and information technology law, with
a focus on legal and regulatory issues surrounding artificial intelligence. Rita has more than
40 peer-reviewed research papers published by leading international publishers. Previously,
she was a legal research fellow at universities in Germany, Japan, Switzerland and Lithuania.
She has prepared research reports for the European Commission, the European Patent Office
and the governments of South Korea and Lithuania, and presented at academic and industry
conferences in Germany, the US, Japan, Hong Kong, Switzerland, Lithuania, Australia and
elsewhere. She is currently Lead Investigator on the projects The Use of Face Recognition
Technologies in Public Sector: Legal Challenges and Possible Solutions and Towards
Explainable AI in Healthcare.
Luke McDonagh is Assistant Professor in Intellectual Property Law at LSE Law School.
Luke holds a PhD from Queen Mary, University of London (2011), an LLM from the London
School of Economics (LSE) (2006–7) and a BCL degree from NUI, Galway (2002–5). He is

xii Research handbook on intellectual property and artificial intelligence
a Fellow of the Higher Education Academy (FHEA). Luke has published widely in respected
journals including Modern Law Review, Journal of Law and Society, Intellectual Property
Quarterly and Civil Justice Quarterly. Luke is the author of the monograph European Patent
Litigation in the Shadow of the Unified Patent Court (Edward Elgar Publishing, 2016) and
the co-author (along with Professor Stavroula Karapapa of the University of Reading) of the
textbook Intellectual Property Law (OUP, 2019). His most recent monograph is Performing
Copyright: Law, Theatre and Authorship (Hart, 2021).
Benjamin Mitra-Kahn is an economist who has served as Economic Advisor to the UK
Intellectual Property Office (2009–12) and Chief Economist at IP Australia (2012–21), and is
currently the programme manager of the Household Expenditure and Income Data Branch at
the Australian Bureau of Statistics. He has published research on the economics of copyright
and patents and was responsible for policy and legislation at IP Australia (2018–21). During
that time the Australian government introduced three IP Bills that, among other things, legis-
lated to allow computerized decision making by an IP office and introduce an Objects clause
to the Patents Act. He holds a PhD in Economics from City University London, an MSc in
Development Economics from the School of Oriental and African Studies, University of
London and a BSc (Hons) in Economics from Royal Holloway, University of London.
Anke Moerland (LLM) is Associate Professor of Intellectual Property Law in the European
and International Law Department, Maastricht University. She holds a PhD on Intellectual
Property Protection in EU Bilateral Trade Agreements from Maastricht University. Her
educational background is in political science (international relations) and law. Anke has
published widely on IP law and policy, with a particular focus on IP law between new tech-
nologies and tradition. In that light, she is at the forefront of discussing the implications of AI
for trade marks, how GI protection can contribute to innovative products that at the same time
preserve tradition, and how copyright rules foster the preservation and digitization of cultural
heritage. Between 2018 and 2020, she held a visiting professorship at Queen Mary University
of London on Intellectual Property Law, Governance and Art. Between 2017 and 2021, Anke
coordinated the EIPIN Innovation Society, a four-year Horizon 2020 grant under the Marie
Skłodowska Curie Action ITN-EJD.
Sean M. O’Connor is Professor of Law at George Mason University, Arlington, VA, USA.
His research and law practice focus on intellectual property and business law, especially the
role of general counsel for start-ups and commercializing innovation in technology and arts.
Professor O’Connor received his law degree from Stanford Law School; a Masters degree
in Philosophy, concentrating on the history and philosophy of science, from Arizona State
University; and a bachelor’s degree in history from the University of Massachusetts, Boston.
Before graduate school he was a singer-songwriter and fronted a rock band with two DIY
albums that received local airplay in the Northeast. He is currently writing The Means of
Innovation: Creation, Control, Methodology and serving as editor of The Oxford Handbook of
Music Law Policy, both forthcoming from Oxford University Press.
Michael D. Pendleton is Emeritus Professor of Law, Murdoch University and Affiliate,
University of Otago. Professor Pendleton has published, taught and practised IP for more
than 40 years. Career highlights include: 12 IP textbooks and practice manuals and more than
100 articles; Chair of the Western Australian Law Reform Commission; Full Professor at the
Law Faculties of the University of Hong Kong, the Chinese University of Hong Kong and

Contributors xiii
City University of Hong Kong; Emeritus Professor of Law at Murdoch University, Australia;
Adjunct Professor of Law, University of Canterbury, Visiting Professor, University of
Auckland, Affiliate, University of Otago; Lawyer, Baker McKenzie, Deacons, Bird Bird
and other international law firms; barrister/solicitor Australia, England and Wales and Hong
Kong; WIPO Mediator, Arbitrator and Domain Name Neutral; Editorial Board, Australian IP
Law Journal; Co-author for 25 years with the late Prof Zheng Chengsi, copyright law drafter
and pioneer of Chinese IP; Life Member, Asian Patent Attorneys Association; and two refer-
ences by the Copyright Law Review Committee (CLRC).
Sharon K. Sandeen is Robins Kaplan LLP Distinguished Professor in Intellectual Property
Law and Director of the IP Institute at Mitchell Hamline School of Law in Saint Paul,
Minnesota. She is a recognized expert on trade secret law, having written (with Elizabeth
Rowe) the first casebook on the subject in the United States. In addition to her books, Professor
Sandeen has written more than 30 articles and book chapters on intellectual property, internet
and information law topics, including detailed analyses of the drafting histories of the Uniform
Trade Secrets Act and Article 39 of the TRIPS Agreement. Professor Sandeen is a member of
the American Law Institute (ALI) and the Association for the Advancement of Teaching and
Research in Intellectual Property (ATRIP), and was the Fulbright-Hanken Distinguished Chair
in Business and Economics for 2019-2020.
Martin Senftleben is Professor of Intellectual Property Law and Director, Institute for
Information Law (IViR), University of Amsterdam. His activities focus on the reconciliation
of private intellectual property rights with competing public interests of a social, cultural or
economic nature. He is President of the Trademark Law Institute (TLI). Professor Senftleben
has provided advice to WIPO in trademark, unfair competition and copyright projects. As
a visiting professor, he has been invited to the National University of Singapore, the Engelberg
Center at NYU Law School, the Oxford Intellectual Property Research Centre, and the
Intellectual Property Research Institute of Xiamen University. His numerous publications
include European Trade Mark Law – A Commentary (with Annette Kur, 2017) and The
Copyright/Trademark Interface (2020). As a guest lecturer, he provides courses at the Centre
for International Intellectual Property Studies (CEIPI), Strasbourg, the Munich Intellectual
Property Law Center (MIPLC), Jagiellonian University Krakow and the University of Catania.
Noam Shemtov joined CCLS in September 2009. He is currently Professor of Intellectual
Property and Technology Law and Deputy Head of the Centre for Commercial Law Studies,
Queen Mary, University of London. He lectures in areas of intellectual property, tech-
nology, and creative industries and his research interests are also focused on these fields.
Professor Shemtov has led research projects and studies funded by UK Research Councils
and by industry, national, supranational and commercial organizations, such as the Arts and
Humanities Research Council (AHRC), CISAC, Microsoft, WIPO, the European Patent
Office, the Foreign and Commonwealth Office (FCO), the UK’s Department for International
Development (DFID) and the European Space Agency. Professor Shemtov also holds visiting
appointments by Spanish and Dutch universities, where he lectures regularly in areas pertain-
ing to intellectual property, creative industries and technology. He is a qualified solicitor both
in the UK and in Israel.
Phoebe Woo, BA (Literary Studies) LLB (HKU) and LLM (Cambridge), is Zue Lo
Pre-Doctoral Fellow, Faculty of Law, University of Hong Kong. She is a co-investigator

xiv Research handbook on intellectual property and artificial intelligence
of two HKU Teaching Development Grant projects and manages the HKU Legal Academy
Student Chapters for knowledge exchange.

2
1. Intellectual property and artificial intelligence:
an introduction
Ryan Abbott
1. INTRODUCTION
In 2020, the AI AlphaFold generated results that won a competition aimed at predicting the
three-dimensional structure of proteins from two-dimensional amino acid sequences. That
competition sounds much less interesting than one involving chess or Go,1
but unlike a board
game, predicting protein folding structure can be a critical element of research and develop-
ment. For example, if you are designing a new antibody to target a protein, or studying diseases
that involve protein misfolding such as Alzheimer’s, it may be necessary to understand what
a protein looks like in three dimensions. DeepMind, which created AlphaFold, more recently
used AI to predict the molecular structure of hundreds of thousands of proteins, including
almost all those in the human body, and it published those results in an open access database.2
AI is not just playing games and doing research; it is doing all sorts of basic activities which
until recently could only be done by a person. These range from taking orders at a restaurant
to driving a car. An AI still cannot safely drive a vehicle fully autonomously, but most of the
time the AI is probably already safer than a human driver. It is likely that pretty soon, AI will
be a better driver than most people all of the time.3
People, after all, set a low bar for driving—
about 94 percent of motor vehicle accidents are caused by human error, which costs more than
a million lives a year worldwide and more than 30,000 lives a year just in the United States.4
If AI, or rather when AI, outperforms average human drivers in the same way it outperforms
people at board games, that will dramatically improve the way we live.
If functionally working for McDonalds or Uber is not impressive enough, AI is performing
tasks that once required highly skilled professionals such as lawyers and doctors.5
The US
Food and Drug Administration has approved AI that autonomously diagnoses certain con-
1
See DeepMind, Alphago, https://deepmind.com/research/case-studies/alphago-the-story-so-far
(last accessed 02/27/2022); see also Sam Byford, Google’s AlphaGo AI defeats world Go number one
KeJie, www.theverge.com/2017/5/23/15679110/go-alphago-ke-jie-match-google-deepmind-ai-2017
(last accessed 02/27/2022).
2
https://alphafold.ebi.ac.uk/. See Jeremy Khan, In giant leap for biology, DeepMind’s A.I.
reveals secret building blocks of human life, https://fortune.com/2021/07/22/deepmind-alphafold-human
-proteome-database-proteins/ (last accessed 02/27/2022).
3
See, e.g., Örebro Universitet, New AI method makes self-driving vehicles better drivers, https://
techxplore.com/news/2021-12-ai-method-self-driving-vehicles-drivers.html (last accessed 02/27/2022).
4
Centre for Disease Control and Prevention, Injury Prevention Control, Global Road Safety,
www.cdc.gov/injury/features/global-road-safety/index.html (last accessed 02/27/2022).
5
Tom Meltzer, Robot doctors, online lawyers and automated architects: the future of the
professions? www.theguardian.com/technology/2014/jun/15/robot-doctors-online-lawyers-automated
-architects-future-professions-jobs-technology (last accessed 02/27/2022).

An introduction 3
ditions such as diabetic retinopathy and macular edema.6
AI is already used extensively in
medical care, but largely as a decision support aid for a human doctor. For example, AI can
analyze a patient’s electrocardiogram and suggest that findings may be consistent with a heart
attack, but ultimately that analysis is just something for a doctor to consider. Only a human
physician can diagnose someone as having a heart attack. However, unlike a decision support
aid, an AI that is autonomously diagnosing a patient is, in a limited context, essentially prac-
ticing medicine. There is evidence that this autonomous AI performs as well as or better than
average doctors.7
Some people are excited by the prospect of robot doctors, and some people are scared—but
before getting scared, realize that human doctors and drivers have some unfortunate similar-
ities. Medical error is one of the leading causes of death.8
To be sure, people are better off
seeing doctors than not seeing doctors, but perhaps in a few years some of the AI that can
autonomously diagnose will do a better job than most doctors. There is some low-hanging fruit
to be had in healthcare.
None of this has escaped the attention of policy makers. In 2021, the UK government
announced a ten-year plan to make the country a global AI superpower and noted that we are
living in “the age of artificial intelligence.”9
The report focuses on investments in planning,
supporting the transition to an AI-enabled economy, and ensuring appropriate governance
frameworks. The United Kingdom was late to the party by comparison to China, which
released its AI strategy in 2017 and boldly planted a flag in the ground. China’s strategy
announced that the nation would become the world leader in AI by 2030, make AI a 30 billion
dollar-a-year domestic industry, and make China a driving force for worldwide ethical norms
and standards. Months later, Russian President Vladimir Putin declared that whoever becomes
the leader in this sphere will become the ruler of the world.10
At least someone took that seriously. According to the United States National Security
Commission on AI, which was formed by Congress in 2018 and which published its report
in March of 2021, China is winning the AI arms race (or at least the United States is falling
behind). It is a 750-page report, but the gist is that the United States is not sufficiently prepared
to defend or compete against China in the AI era.11
It proposes that America increases spend-
ing on non-defense AI RD all the way up to 32 billion dollars a year in 2026—about what
America now spends on biomedical research. The report also has a chapter on IP and notes
that IP rules are important for national defense as well as industrial strategy. It even highlights
some of the specific issues that need to be addressed for the United States to be successful at
using and promoting AI—including my personal favorite, AI inventorship.
6
FDA, FDA permits marketing of artificial intelligence-based device to detect certain
diabetes-related eye problems, www.fda.gov/news-events/press-announcements/fda-permits-marketing
-artificial-intelligence-based-device-detect-certain-diabetes-related-eye (last accessed 02/27/2022).
7
Ibid.
8
For the landmark report on the subject, see Institute of Medicine. 2000. To Err Is Human: Building
a Safer Health System. Washington, DC: The National Academies Press. https://
doi
.org/
10
.17226/
9728.
9
Government of the United Kingdom, National AI Strategy, www.gov.uk/government/publications/
national
-ai
-strategy (last accessed 02/27/2022).
10
David Meyer, Vladimir Putin says whoever leads in artificial intelligence will rule the world,
https://fortune.com/2017/09/04/ai-artificial-intelligence-putin-rule-world/ (last accessed 02/27/2022).
11
NSCAI, Final Report, National Security Commission on Artificial Intelligence, www.nscai.gov/
wp-content/uploads/2021/03/Full-Report-Digital-1.pdf (last accessed 02/27/2022).

4 Research handbook on intellectual property and artificial intelligence
Different jurisdictions are not just competing in terms of technology and investment, they
are also competing with respect to governance frameworks for AI. Having appropriate laws
in place is not only a critical component of industrial strategy and economic competitiveness;
more broadly, it is vital to ensuring that technology such as AI is developed and used in
ways that promote social value and that limit harm. All technologies and activities can cause
harm, and AI is no exception. AI has been used for political and social manipulation, it has
caused flash crashes in the stock market, and some fairly prominent technologists think it may
doom the human race.12
Governance is not all talk and reports—the European Commission
recently came out with draft AI-specific regulations which propose a risk-based approach
that completely prohibits certain uses of AI, such as for indiscriminate facial recognition in
public places (with some exceptions) and for generating social scores of citizens.13
It proposes
a variety of compliance mechanisms for high-risk AI systems and uses, and it takes a lighter
approach to lower or minimal risk systems. This particular set of regulations does not focus
directly on AI and IP, but AI and IP is a key part of the ongoing debate about appropriate legal
frameworks. Just as rules that take AI into account are important generally so that we get the
best possible outcomes, they are important specifically with respect to IP to make sure that
IP systems can achieve their underlying goals. AI is going to be broadly economically and
socially disruptive, and specifically disruptive to IP systems.
AI has been around for a long time, and to some extent so have the challenges it poses to
IP rules.14
For instance, since the 1960s people have claimed to have developed computers
capable of making music. In 1966 the United States Register of Copyright publicly ques-
tioned whether these works should be copyrightable and disclosed that the office had already
received applications for at least partially computer-generated works including musical com-
positions.15
But that music was all terrible. Today’s AI has come a long way since the 1960s,
and OpenAI’s music is now practically mediocre. The project’s website has AI that makes
music in the style of current and recently deceased artists, which raises another interesting set
of IP issues.16
As AI-generated music improves it is worth keeping in mind that it is only going to get
better. At some point, we are going to have AI that can make music people want to hear. That
is going to be incredibly disruptive because once you have AI that can make good music, you
can have AI make a virtually unlimited amount of it at almost no marginal cost. That is going
to be competition for human musicians in the way that self-driving Ubers are going to be
competition for human drivers. But AI will not just replace people; it will change the way that
12
See, e.g., Maureen Dowd, Elon Musk’s billion dollar crusade to stop the A.I. apocalypse,
www.vanityfair.com/news/2017/03/elon-musk-billion-dollar-crusade-to-stop-ai-space-x (last
accessed 02/27/22).
13
European Commission, Proposal for a Regulation of the European Parliament and of the Council
Laying Down Harmonised Rules on Artificial Intelligence (Artificial Intelligence Act) and Amending
Certain Union Legislative Acts COM/2021/206 final, https://eur-lex.europa.eu/legal-content/EN/TXT/
?uri
=
CELEX
%3A52021PC0206 (last accessed 02/27/2022). I think the concern is not really with AI
being used to make social scores but more with the idea of social scores per se, but anyway.
14
Thomas Ling, AI is about to shake up music forever—but not in the way you think, www
.sciencefocus.com/science/ai-music-future/#:~:text=How%20easy%20is%20it%20to,by%20repeatedly
%20being
%20shown
%20things (last accessed 02/27/2022).
15
See U.S. COPYRIGHT OFFICE, SIXTY-EIGHTH ANN. REP. REG. COPYRIGHTS 4–5 (1966).
16
https://openai.com/blog/jukebox/.

An introduction 5
music is made and consumed. AI will make personalized music for someone in real time, to
improve their mood or accompany their workout.
Harold Cohen and his AI system AARON were making paintings in the 1970s that, for
all I know, could have been made by an AI, one of the great modern artists of the twentieth
century, or one of my toddlers.17
That probably says more about my lack of artistic sense than
anything to do with AI, but as with AI musicians, the past few decades have been good for AI
painters. Or, at least, their art is selling for more. Headlines were made in 2018 when Christie’s
became the first major auction house to offer an AI-generated artwork for sale. It went for
$432,500 dollars.18
Now exhibits of AI-generated artwork are no longer remarkable.
If you are looking for shocking headlines in the art world, it is probably what people are
paying for Non-Fungible Tokens (NFT). An NFT is basically data stored on a digital ledger
that can represent things such as a photograph or a painting.19
Some of the NFTs now being
exchanged for staggering sums of money were AI-generated. AI making art is going to be
socially and economically disruptive just like AI making music, and it will also result in more
business for copyright attorneys. Normally when a person makes a digital painting the work
automatically attracts copyright protection more or less worldwide. That prevents someone
else from copying or selling the work for the author’s lifetime plus, in the United States and
most large markets, 70 years. What if that author is an AI?20
In fact, numerous legal questions are raised by these sorts of technological advances. With
respect to AI generating historically copyrightable work: Can you get a copyright for an AI
generated work? Who would own it? How long would copyright exist? Until recently these
were basically academic questions because there was not much of a market for AI-generated
content, but that is changing quickly given the speed at which AI is improving. These are,
or will soon be, now commercially important questions. Similar considerations apply to AI
making other sorts of IP historically protectable by patents, designs, and trademarks. For
instance, AI is making trademarks,21
but perhaps more interesting is how AI can fundamentally
alter how trademarks function. Trademarks are designed to indicate to consumers the source
and origin of goods and services. What happens when commerce is online, and when AI is
buying goods on your behalf? How does Alexa determine the source and origin of goods and
services and what sort of competing signs would Alexa find confusing or deceptive? How do
we apply infringement tests based on human consumers to AI-based purchases? What happens
where AI is effectively both the buyer and the seller?
At the end of 2019, the United States Patent and Trademark Office (USPTO) put out
a request for comments on AI and patents as well as a request for comments on AI and IP
more generally.22
It published a report summarizing certain comments it received early last
17
Chris Garcia, Hardo Cohen and AARON—A 40-year collaboration, https://computerhistory.org/
blog/harold-cohen-and-aaron-a-40-year-collaboration/.
18
Christie’s, Is artificial intelligence set to become art’s next medium, www.christies.com/features/a
-collaboration-between-two-artists-one-human-one-a-machine-9332-1.aspx (last accessed 02/27/2022).
19
Non-Fungible Tokens (NFT), https://ethereum.org/en/nft/ (last accessed 02/27/2022).
20
Ahmed Elgammal, AI is blurring the definition of artist, www.americanscientist.org/article/ai-is
-blurring-the-definition-of-artist#:~:text=To%20create%20AI%20art%2C%20artists,the%20aesthetics
%20it
%20has
%20learned (last accessed 02/27/2022).
21
See, e.g., www.brandmark.io.
22
USPTO, Request for Comments on Intellectual Property Protection for Artificial Intelligence
Innovation (Federal Registry Notice 84 FR 58141), www
.uspto
.gov/
sites/
default/
files/
documents/
ITIF_RFC-84-FR-58141.pdf (last accessed 02/27/2022).

year.23
The United Kingdom Intellectual Property Office launched a similar consultation to the
USPTO in 2020 and published its results in 2021.24
One of the most interesting outcomes was
how different the UKIPO findings were from the USPTO findings—even with many of the
same stakeholders providing submissions. Part of the difference may be the result of rapidly
changing attitudes toward AI and IP. Again, these consultations are not just about AI making
new IP—they are also about how AI is being developed using IP. For that matter, AI itself can
be protected by IP rights.
On top of which, AI uses IP. Consider Clearview’s AI, which has collected and trained
on more than ten billion images from across the Internet. Clearview licenses this AI to law
enforcement to use for facial recognition.25
Usually copying an image constitutes copyright
infringement, but what if it is being copied to train an AI, or just to generate insights? Is that
infringement? What if a bunch of machines are sending digital copies of images between each
other and no one ever sees those copies? What if an AI is functionally infringing someone
else’s intellectual property but not in a way that is directly attributable to the person? Who
would be responsible for that and on what basis?
Different jurisdictions already have different answers to some of these issues. For instance,
there is a very broad exception to copyright infringement in Japan for text and data mining
and a much narrower one in Europe.26
But there is not yet much case law exploring in what
situations it would be infringement for a machine to do something that would be infringement
for a person. As different jurisdictions come up with different answers to these questions
another challenge will be presented because of the transnational nature of IP. A book that you
write in the United States is protected in France, Brazil, and India. That is not now the case
with a book written “by” an AI. World Intellectual Property Organization (WIPO), the UN
agency most responsible for IP, has gotten in on the action by hosting a series of stakeholder
discussions around IP and AI (recently broadened to IP and frontier technologies).27
In theory,
an international treaty may come out of this work so that jurisdictions around the world can
agree to a common set of rules involving IP and AI—but that is a slow process. Not oblivious
to the time frames involved, WIPO is working on framing the debate and getting stakeholders
to ask the right questions rather than proposing specific substantive solutions.
Getting consensus on definitions and language would be a good start, as it is remarkable
that about 65 years after the term AI was coined it still lacks a widely accepted definition.
Definitional problems are significant when laws are being written to regulate AI and some
of the people debating these laws are talking about different things. If people cannot agree
23
US Patent and Trademark Office, Public Views on Artificial Intelligence and Intellectual Property
Policy. Oct 2020. www.uspto.gov/sites/default/files/documents/USPTO_AI-Report_2020-10-07.pdf.
24
UK Intellectual Property Office, Government Response to Call for Views on Artificial Intelligence
and Intellectual Property. Updated 23 March 2021. www.gov.uk/government/consultations/artificial
-intelligence-and-intellectual-property-call-for-views/government-response-to-call-for-views-on
-artificial-intelligence-and-intellectual-property.
25
Kashmir Hill, The secretive company that might end privacy as we know it, www.nytimes.com/
2020/01/18/technology/clearview-privacy-facial-recognition.html (last accessed 02/27/2022).
26
European Alliance for Research Excellence, Japan amends its copyright legislation to meet future
demands in AI and Big Data, https://eare.eu/japan-amends-tdm-exception-copyright/ (last accessed
02/27/2022).
27
WIPO, The WIPO Conversation on Intellectual Property and Frontier Technologies www.wipo
.int/about-ip/en/frontier_technologies/frontier_conversation.html (last accessed 02/27/2022).

An introduction 7
on what AI is, it will be hard for them to agree on the definition for an AI-generated work.
Without a common understanding of an AI-generated work it becomes difficult to have har-
monized IP rules.
Some of the WIPO conversations have focused on issues that have been debated for decades,
such as whether you can patent software and whether someone should be able to “own” data.
These are not new issues, but they are issues that are taking on newfound importance due to
AI. Patents on new software architectures and uses may be more valuable now that AI can do
more, and data ownership matters more given the importance of data for training AI, for AI
using data to generate insights, and for AI generating new sorts of data. Some of these issues
are more specific to advances in AI, such as deepfakes. Deepfakes are not new, in the sense
that there have long been technologies that can be used for deception, but what is new is how
easy it has become for just about anyone to use AI to make a realistic-appearing video. The law
has long dealt with deception, but the evolution of technology means that some of our rules
may need to be rethought.
People are also interested in AI in administrative decision making. Not only can AI make
new IP or help people to make IP, but it can also file or help people to file for IP rights. For
example, an application called Specifio can generate a patent application based on a single
claim.28
It does not directly replace a patent attorney because the AI is not as good as
a person, and a patent attorney needs to (or at least should) review and revise what the AI is
generating—but if it takes that attorney half as much time to write an application with the help
of the AI, then perhaps we only need half as many patent attorneys (or, at least, fewer junior
attorneys). Perhaps we will still need all the lawyers but they will be more productive so we
will have more IP filings. That suggests that IP offices may get busier and need to turn to AI
for help. In fact, the USPTO is already using AI to classify incoming patent applications and
determine who should review them. It may not be long before a patent application arrives at
a patent office and AI classifies it, does a first pass at a prior art search, and then drafts an
office action for a human examiner to review. Even without an AI replacing anyone, that is
going to change the process of patent prosecution in all sorts of subtle and not-so-subtle ways.
The public is concerned about AI in decision making—not really with patent prosecution,
because it is hard for the public to care about patent prosecution even with robots involved,
but certainly with AI augmenting and automating administrative and even judicial decisions
in other contexts. In part this is because of worries around impermissible biases and explain-
ability and transparency. It would be bad to have AI making recommendations at criminal
hearings without being able to explain the basis for those recommendations, particularly if its
recommendations were based on protected characteristics.29
2. ADVANCING THE CONVERSATION
While I do not personally have the answer to every question involving IP and AI, this
Handbook goes a long way toward advancing our understanding of many of the most impor-
28
Auto-generated Software-patent applications, https://
specif
.io/(last accessed 02/27/2022).
29
Melissa Hamilton, We use Big Data to sentence criminals. But can the algorithms really
tell us what we need to know? https://openresearch.surrey.ac.uk/permalink/44SUR_INST/15d8lgh/
alma99516814902346 (last accessed 02/27/2022).

tant ones. A group of world-leading authorities from a diverse group of jurisdictions, disci-
plines, and professional fields have generously contributed original content here.
The Handbook is divided into four parts. Part I includes contributions, such as the present
chapter, that address intellectual property rights (IPRs) generally rather than focusing primar-
ily on a specific type of IPR, such as trademarks or patents. Of course, even those chapters
that focus on a particular type of IPR have broader relevance as many of their insights apply in
other contexts—including outside of IP law.
Following this chapter, Gervais argues that AI-generated content should not be treated the
same way in patent and copyright law because the two bodies of law have different goals.
He is particularly wary of conflating human and AI activity in the copyright sphere, given its
potential to disadvantage human creatives. Gervais suggests “IP proximate cause” could be
used to determine where a person has contributed to an invention or a copyrightable work and
therefore whether protection should be granted. Barfield, Karanasiou, and Chagnal-Feferkorn
then explore the concept of “embodiment” in AI and its relevance to IP law. They argue
that embodiment, either physically (as with a robot) or virtually (as with an avatar in virtual
reality), is an important concept for IP because embodied agents can be both the subject of
IPRs and infringers of IPRs. They explain how the lens of embodiment sheds light on the
wider socio-legal narratives in AI and IP.
O’Connor considers artist “style,” which has limited protection under copyright, trademark,
and right of publicity laws. Yet AI’s ability to mimic style suggests that style is important,
including for its newfound commercial relevance for AI, and also that style is capable of being
quantified and fixed and thus should be entitled to greater protection. Deltorn considers infer-
ence models, which are complex entities combining know-how to produce a relevant solution
with information derived from machine training data. These models underlie many of the most
impressive recent AI breakthroughs, and Deltorn considers how such models can be protected
under European frameworks by various IPRs including copyright, database protections, trade
secrets, and patents. Finally, Pendleton provides a skeptical take on the fitness of existing IP
frameworks, arguing that for IP law to be useful for AI it should focus on misappropriation.
Such an approach should protect investment, whether financial or based on labor or skill, but
should also provide that subsequent investments which result in a transformative use should
protect against infringement and give rise to new rights.
Part II focuses on copyright and related rights. Several contributors present compelling nar-
ratives around AI authorship and AI-generated works, starting with Craig, who, like Gervais,
suggests that advances in AI give us the opportunity to rethink the purposes of copyright law.
She argues that the principles that underlie copyright law should inform our response to AI,
and that we should attempt to achieve normative equilibrium when confronted by disrup-
tive technologies. On this basis, AI-generated outputs should not be eligible for copyright
protection, and use of protected material for training AI should be non-infringing. Frosio
subsequently considers AI authorship, initially by considering whether AI meets traditional
authorship requirements, then by considering various approaches to AI-generated works
from a policy perspective. He proposes either that AI-generated works could go to the public
domain, a person or a machine could be grated authorship, the work could receive specialized
sui generis protection, or rights could be granted to publishers and disseminators. Frosio
weighs the costs and benefits of these options from different theoretical perspectives.
Lee and Woo argue that completely overhauling copyright law in response to AI is not
optimal, and that simple tests of exploitation, while appealing, are not an adequate substitute

An introduction 9
for a multi-layer regime. They defend the current copyright framework based on examples
from European and Asian case law. They conclude that top-down solutions are not sufficient
for addressing complex copyright challenges and that non-statutory approaches are needed.
After this, Benhamou and Andrijevic examine use of AI in a more specific context: generation
of new pictures. They analyze the criterion of originality under copyright law across the full
lifecycle of an AI-generated picture from the input to the output stage. They conclude that
copyright law protects elements of this process at numerous stages and in complex ways, and
that authorship and entitlement associated with AI output will often need to be assessed on
a case-by-case basis. Justice Arnold turns to AI and performers’ rights, which is a statutory
right under English law, but international legal frameworks also require certain protections
for performers. He considers whether performers’ rights protect, or should protect, real per-
formers against imitation by avatars and deepfakes, and whether performers’ rights protect, or
should protect, imaginary avatars.
Ghosh argues that AI is simply software and that recognizing this allows a developed body
of law related to software and copyright to be applied to activities involving AI. It also con-
nects IP and AI to computer science as an academic discipline with typologies that are useful
for assessing legal questions of ownership, liability, and transactions. Ghosh explains how
computer science can inform debates over copyright, with parallels to patent law, and argues
it offers an explanation for the 2021 US Supreme Court decision in Oracle v Google. Finally,
Bonadio, Dinev, and McDonagh consider use of IP to train AI as well as use of IP by AI for
activities such as text and data mining. They consider whether such activities should constitute
infringement, whether exceptions should apply and in what circumstances, and who should be
liable for infringement. They explore whether AI-based activity should receive more gener-
ous fair use or fair dealing treatment with a focus on the United States, United Kingdom, and
European Union.
Part III focuses on trademarks and design rights. Lim begins by exploring AI’s impact
on likelihood of confusion as the test for trademark infringement. He considers how AI will
change how consumers purchase goods and services, liability for sales of counterfeit products
on platforms such as Amazon, and even how AI can facilitate adjudication of trademark
disputes. Next, Moerland and Kafrouni consider trademarks in the context of AI-based pur-
chasing and argue that for trademarks to continue to fulfill their roles, the average consumer
may need to become the AI-assisted consumer. They explain that improved understanding of
online purchasing behavior and AI use is needed to properly apply a test based on consumer
confusion in this new environment.
Senftleben examines the rules proposed by the European Commission in its Proposal for
a Digital Services Act to create new transparency obligations for accountable digital services
which focus on transparency measures for consumers. Exploring AI-driven behavioral adver-
tising practices and keyword advertising cases, he explains that rights holders also have an
interest in advertising parameters and that information on market alternatives could improve
trust in AI-based personalized advertising. Gangee continues the exploration of trademark law
by outlining the ways in which AI is being used in the registration environment, posing some
non-obvious doctrinal questions. He next surveys AI being used for enforcement, and argues
that technology is shifting goal posts and setting new defaults. Last, Cook turns to design law
and AI-generated designs. He proposes that design law may be better understood as a hybrid
of patent and copyright law rather than an independent area of law, and that this helps us to
understand whether and how we should provide protection for AI-generated designs.

Part IV focuses on patents and trade secrets. Crouch begins by considering another form
of AI, namely corporations and other private organizations. He notes that corporations have
already obtained legal personhood and certain civil rights, and he argues there are numerous
parallels between corporations and AI. He traces the progression of corporate rights in the
patent context from pre-industrial times to the America Invents Act of 2011. He concludes
that the patent system can effectively function, and AI-generated inventions can be protected,
without having to rely excessively on legal fictions. Mitra-Kahnv analyzes economic reasons
for recognizing AI patent inventors. He claims there is no convincing economic reason to
allow AI inventorship to incentivize innovation but some reason for doing so to encourage
technology transfer, and he proposes some options to protect AI-generated inventions without
AI inventorship.
Bayern considers AI’s impact on trade secret law and policy and notes that AI can dramat-
ically reduce costs associated with reverse engineering. He also notes that AI may be par-
ticularly susceptible to reverse engineering, even AI systems with limited explainability and
transparency. Bayern explores these ideas and suggests they may have significant impact on
trade secret policy and that trade secret law may need to change. Matulionyte and Aranovich
then contemplate the tension between the desire for explainable AI and trade secret protection.
They consider this tension in depth in one case study in judicial administration and one in
healthcare, and suggest solutions to areas of conflict.
Shemtov and Gabison consider the impact of AI on inventive-step requirements for patent-
ability. They note that the person skilled in the art will need to be augmented as AI becomes
more common—the nature of the augmentation being a question of fact. This is also justified
under incentive theory, and the threshold for patentability will need to increase as AI makes
invention easier so that the patent system can continue to achieve its goals. In the final chapter,
Sandeen and Aplin interrogate the extent to which machine-generated data and algorithms
may qualify as trade secrets under US and EU law. They show, through case studies involving
autonomous vehicles and credit scoring, that commonly claimed subject matter of trade secret
protection may not actually qualify for protection. However, factual secrecy in many cases
provides significant protection even in the absence of trade secret protection, and so regulatory
efforts may need to focus on factually secret data.
Sandeen and Aplin interrogate the extent to which machine-generated data and algorithms
may qualify as trade secrets under US and EU law. They show, though case studies involving
autonomous vehicles and credit scoring, that commonly claimed subject matter of trade secret
protection may not actually qualify for protection. However, factual secrecy in many cases
provides significant protection even in the absence of trade secret protection, and so regulatory
efforts may need to focus on factually secret data. In the final chapter, Shemtov and Gabison
consider the impact of AI on inventive-step requirements for patentability. They note that the
person skilled in the art will need to be augmented as AI becomes more common—the nature
of the augmentation being a question of fact. This is also justified under incentive theory, and
the threshold for patentability will need to increase as AI makes invention easier for the patent
system to achieve its goals.

An introduction 11
3. AI-GENERATED WORKS AND THE ARTIFICIAL
INVENTOR PROJECT
My own AI and IP research has focused on AI-generated works, mainly in the context of
patent and copyright law. A recent case study by Siemens provides a good introduction to the
topic.30
Siemens developed a new car suspension which it wanted to patent, but determined
that was not possible because a human inventor could not be identified for a patent application.
Essentially, the human engineers involved stated that the design was generated by an AI and
that none of them had done anything to qualify as inventors. Getting inventorship right is more
important in some jurisdictions than others, but in the United States it is a criminal offense
to deliberately inaccurately list yourself as an inventor and failing to identify all inventors in
good faith can render a patent invalid or unenforceable.31
Without an inventor, conventional
wisdom holds, an invention cannot be patented.
The takeaway from this case is that Siemens has some nifty AI which is functionally
stepping into the shoes of employees and performing tasks that used to make a person an
inventor. But, because the law in most jurisdictions treats behavior by a person and behavior
by a machine differently, Siemens could not protect its invention in the way it could if it had
been made by one of its human employees. This is the most commercially relevant challenge
with AI-generated inventions—subsistence of patent rights. It also raises questions of who, or
what, would be listed as an inventor on a patent application for an AI-generated invention, and
who, or what, would own that invention.
As mentioned above, the terminology for AI doing various things in the IP context is
all over the place—terms such as AI inventions,32
AI art, AI-assisted or AI-generated, and
computer-generated abound. Aside from the fact that IP offices, courts, policy makers, and
even the various contributors to this book use different terms, different people can also use
the same term to mean different things. The rest of the chapter will focus on “AI-generated
works,” which I will define as an AI output for which the AI has functionally acted as
a traditional human author or inventor. Sometimes AI-generated works are defined as works
lacking a traditional human author, but I think that approach is problematic in cases where
a person and a machine independently make contributions that would qualify people for joint
authorship or inventorship. As a result, the contribution and financial interests of the party
contributing the AI are ignored.
While on definitions, it is hard to talk about AI-generated inventions without defining AI.
I define it as “an algorithm or machine capable of completing tasks that would otherwise
30
Siemens, Re; Draft Issues Paper on Intellectual Property Policy and Artificial Intelligence, www
.wipo.int/export/sites/www/about-ip/en/artificial_intelligence/call_for_comments/pdf/org_siemens.pdf
31
37 CFR 1.56 Duty to disclose information material to patentability (“[N]o patent will be granted
on an application in connection with which fraud on the Office was practiced or attempted or the duty of
disclosure was violated through bad faith or intentional misconduct”).
32
www.federalregister.gov/documents/2019/08/27/2019-18443/request-for-comments-on-patenting
-artificial
-intelligence
-inventions. I think AI Inventions as a term is unfortunately ambiguous, but in
fairness my attempt to brand AI-generated inventions as “computational inventions” was wildly unsuc-
cessful. Ryan Abbott, I Think, Therefore I Invent: Creative Computers and the Future of Patent Law, 57
B.C. L. Rev. 1079 (2016).

require cognition.”33
There are many ways to build AI, but with particularly with respect to the
activities considered here, as Ghosh explores in depth and as a practical matter, AI is more or
less software.34
The claim that AI can invent is still somewhat controversial, although much less so than
it was even a few years ago. However, it is not a controversial statement to say that AI is
generating output that would traditionally get copyright protection without someone who
traditionally qualifies as an author. That is because the bar for making something that gets
copyright is quite low. Taking a photograph of this page with your smartphone would make
you an author of the photograph. It would be difficult to claim that an AI cannot engage in that
sort of activity—even a monkey can do it. Your smartphone photo is not a great work of art,
but copyright law does not concern itself with whether you’ve made the next Mona Lisa or
something that only Clearview’s AI will ever see. Both works will be protected by copyright
long after you are dead. But creating something patentable is a lot harder than coming up with
something protectable by copyright. For an invention to be patentable, it has to be: (1) new,
so that no one in human history has ever disclosed it; (2) non-obvious to a skilled person who
essentially represents an average researcher in your field; and (3) useful.
People routinely use AI in the process of inventing without AI inventorship being an issue.
That is because there is a test for what makes an inventor—whether the inventor, at least under
US law, “conceived” of an invention; whether they had the “definite and permanent idea of an
operative invention, including every feature of the subject matter sought to be patented.”35
So
an AI, or another person, can do quite a bit of heavy lifting with an invention before questions
of inventorship start to arise. Plus, if at least one person qualifies as an inventor, then at least
someone is getting a patent (and who that person is can be altered by contract), so the subsist-
ence problem does not arise.
Can a machine conceive of something? Some people believe that because we do not have
machines that think the way that people do, a machine cannot invent. It is certainly true that
machines do not think the same way that people do—no robot ever woke up in the morning,
had an existential crisis, and decided it was going to start inventing instead of assembling cars
on a manufacturing line. But so what? AI can be given known problems, such as finding the
design for a more effective car suspension, and can generate a solution that is new and inven-
tive to an average researcher, and can then present that solution so that anyone who under-
stands car suspension designs would recognize its utility. That means the AI is generating the
entire idea of an operative invention without a person doing anything that should qualify them
to be an inventor.
Sometimes even with AI doing a lot of the work, a person could still qualify as an inventor.
For example, occasionally the tricky bit of invention is finding a problem to be solved. Where
someone has exercised inventive skill to formulate a problem, or perhaps even to formulate
a problem in a manner that would be understood by an AI, that could make them an inventor.
But that is not usually how invention works. Most of the time people are trying to solve known
33
Ryan Abbott, The Reasonable Robot (2020) at 22.
34
There is some special purpose hardware that one could consider AI, and fancy issues when the AI
is operating on a distributed leger or as open source code, but we needed a contributor to explore those
issues.
35
Sewall, 21 F.3d at 415.

An introduction 13
problems—designing a better toothbrush, creating a new antibody to treat COVID-19,36
and
so on. Sometimes a person programming or designing or training an AI could qualify as an
inventor. That seems appropriate at least in cases where someone is designing an AI to solve
a specific problem, and where they know the problem being solved and have an expectation
their AI will generate a solution. It seems less appropriate in cases where someone is building
an AI to solve problems at a more general level, such as optimizing industrial components,
without knowing whether some other team is going to apply that AI to optimizing a computer
chip or an antennae design. For that matter, dozens or hundreds of programmers may contrib-
ute to an AI, and AI may be, in whole or part, open-source code.
Sometimes a person recognizing the value of an AI’s output can qualify as an inventor.
That works when someone is the first to notice that penicillin is inhibiting bacterial growth,
or that there might be a use for Viagra outside of treating heart disease. It also works if an AI
suggests a dozen possible antibodies to treat COVID-19 and human researchers then need
to do additional testing to find the best one. But if an AI can validate its own output—if it
can generate a million possible car suspension designs, model them, rank them according to
generally known criteria for function, and then essentially say “here is the best car suspension
out of a million and it dramatically outperforms what we have in our vehicles today”—then
attributing inventorship to the person who says “right, let’s go with the option the AI likes”
does not seem right. At least some of the time what we think about as inventive activity is
being done, at least functionally, by the AI and not by a person—regardless of whether people
are just very complex biological machines or whether what neural networks do analogizes
well to what goes on in someone’s head. Patent law should focus on function because society
benefits from functional behavior—the result of AI inventors is more innovation, more open
knowledge, and more commercially valuable products.
A couple of years ago there was no case law on AI-generated inventions. There were juris-
dictions that required an inventor to be a natural person, but whether by statute or case law,
the possibility of a non-human inventor was only something that had been considered in the
context of corporate inventorship.37
Whether a company could be an inventor is an interesting
question—particularly considering legal theories that suggest a company is more than a sum
of its individual agents, a theory that supports criminal liability directly for companies—but
in any event, it presents a very different set of questions from AI inventorship. Companies act
through human agents, and if companies did not have to list inventors on patent applications
those inventors would fail to receive due credit and sometimes financial rewards. But in the
case of AI-generated inventions, there is no human inventor who is being denied credit. To the
contrary, allowing someone to list herself on a patent application for an AI-generated inven-
tion would reduce transparency and allow someone to receive false credit.
There was some law with respect to AI-generated works and copyright protection. The
United Kingdom was the first country to explicitly provide copyright protection for so called
36
AI has already played a key role in RD related to COVID-19. Hannah Kuchler, Will AI turbo-
charge the hunt for new drugs? Financial Times, March 19, 2022, www
.ft
.com/
content/
3e57ad6c
-493d
-4874-a663-0cb200d3cdb5.
37
See, e.g., Univ. of Utah v Max-Planck-Gesellschaft zur Forderung der Wissenschaften e.V. 734
F.3d 1315 (Fed. Cir. 2013).

computer-generated works back in 1988.38
Those are works generated by a computer in such
circumstances that there is no human author of the work.39
In such cases, the producer of the
work, the person who undertakes to have the work created, is deemed to be the author and the
work receives a shortened period of statutory protection—50 years versus the usual lifetime of
an author plus 70 years. The United Kingdom remains an international outlier in this respect,
although some Commonwealth countries followed suit.40
There has only ever been one case in the United Kingdom involving copyright infringement
and a computer-generated work under this law, but even then, no one was challenging the cop-
yright in the underlying work.41
There are a few reasons why there has not been more litigation
over AI-generated works. First, unlike the United States, you cannot register copyright in the
United Kingdom, so there would only be litigation involving infringement. Second, there is
not that much copyright litigation in the United Kingdom where the subsistence of copyright
is at issue. Third, even if someone was aware that a work was AI-generated, there would be no
point challenging subsistence given the state of UK law. Finally, and I think most importantly,
until recently AI-generated works have not had much commercial value and thus have not
been worth suing over. That has changed and will continue to change now that AI is getting
better at making creative works.42
By contrast to the United Kingdom, the United States Copyright Office has a “Human
Authorship Requirement” that has been formally in place since 1973. AI-generated works
automatically enter the public domain once disclosed and cannot receive copyright protection.
That means if you have an AI that makes a song that becomes a hit single, it will be awfully
tempting to take credit for that work. The machine is unlikely to complain. The Copyright
Office cites to the 1884 case of Burrow Giles v Sarony in support of this policy.43
That is the
Supreme Court case that first held that a photograph could be protected by copyright. Napoleon
Sarony had sued the Burrow Giles Lithographic Company for copyright infringement of his
famous photograph of Oscar Wilde, and the company alleged that the photograph could not be
protected because it was merely “a reproduction on paper of the exact features of some natural
object or of some person.”44
But the Supreme Court held that “writings” by which “ideas in the
mind of the author are given visible expression” were eligible for protection, and it referred to
authors as human.45
The Copyright Office interprets this case to hold that copyright requires
38
The Copyright, Designs and Patents Act 1998 (CDPA), Section 9(3) (“In the case of a literary,
dramatic, musical or artistic work which is computer-generated, the author shall be taken to be the person
by whom the arrangement necessary for the creation of the work are undertaken”). Other jurisdictions
have adopted a similar rule including Ireland, Hong Kong, New Zealand, and South Africa.
39
Ibid.
40
E.g., Copyright Act of 1994, s 5 (NZ); Copyright and Related Rights Act 2000, pt 1, s 2 (No
28/2000) (IR).
41
Nova Production v Mazooma Games, EWHC 24 (Ch) [2006].
42
Ian Bogost, The AI–art gold rush is here. The Atlantic, March 6, 2019, www
.theatlantic
.com/technology/archive/2019/03/ai-created-art-invades-chelsea-gallery-scene/584134/ (last accessed
02/27/2022); AIArtist.org, 41 creative tools to generate AI art, https://aiartists.org/ai-generated-art-tools
43
Burrow-Giles Lithographic Co. v Sarony, 111 U.S. 53 (1884).
44
Ibid at 56.
45
Ibid at 58.

An introduction 15
creative powers of the mind which it thinks machines lack.46
There has never been a US case
where a court has considered whether an AI-generated work could be protected or whether an
AI could be an author.47
That Copyright Office policy was almost challenged in court a few years ago.48
The case
involved a series of pictures that a macaque named Naruto took of himself using photographic
equipment belonging to a nature photographer, David Slater.49
The photographs ended up
having value because people thought a smiling monkey selfie was cute and entertaining even
though black crested macaques smile as a display of aggression. So Naruto is likely respond-
ing to his own reflection in the camera lens and attempting to intimidate that monkey. In any
case, people started using the photograph without Slater’s permission. He accused them of
copyright infringement, and this all resulted in enough attention that the Copyright Office
reformulated its Human Authorship Requirement in its Compendium and explicitly listed
photographs taken by a monkey as ineligible for protection.50
This seemed to put an end to the matter until People for the Ethical Treatment of Animals
(PETA) sued Slater in Federal Court alleging that Naruto was the author and owner of the
photographs, and that Slater was liable for copyright infringement.51
The case was ultimately
dismissed by the Ninth Circuit Court of Appeals, but it was dismissed based on standing. The
Courts said that unless the Copyright Act plainly states that non-human animals have standing
to sue, we are not going to let them bring lawsuits. As a result, the Copyright Office policy has
never been challenged.52
What about in the patent context? Until recently, conventional wisdom held that if you
did not have a human inventor then you could not get a patent. But there are reasons to think
that is not a good approach. The patent system exists to incentivize invention, to encourage
disclosure of inventions that might otherwise be kept as trade secrets, and to encourage the
commercialization of inventions. Historically, some academics have thought machines do not
care about patents so there would be no point to patenting their inventions. Of course, it is true
that AI does not care about patents, but companies like Siemens and Pfizer care about patents.
If we are moving from a paradigm of research and development in which we want to directly
incentivize people to invent to one in which we want to incentivize people to build AI that will
invent, then protecting the output of that AI is critical to promoting that activity.
46
Compendium of the US Copyright Office Practices, Third Edition, Section 306 (“Human
Authorship Requirement. The copyright law only protects ‘the fruits of intellectual labor’ that ‘are
founded in the creative powers of the mind.’ Trade-Mark Cases, 100 U.S. 82, 94 (1879). Because copy-
right law is limited to ‘original intellectual conceptions of the author,’ the Office will refuse to register
a claim if it determines that a human being did not create the work. Burrow-Giles Lithographic Co. v.
Sarony, 111 U.S. 53, 58 (1884)”) .
47
Second Request for Reconsideration for Refusal to Register A Recent Entrance to Paradise,
February 14, 2022, www.copyright.gov/rulings-filings/review-board/docs/a-recent-entrance-to-paradise
.pdf (last accessed 02/27/2022).
48
Naruto v Slater, 888 F.3d 418 (9th Cir. 2018).
49
Andres Guadamuz, Can the monkey selfie case teach us anything about copyright law? www.wipo
.int/wipo_magazine/en/2018/01/article_0007.html (last accessed 02/27/2022).
50
Compendium of the US Copyright Office Practices, Third Edition, Section 313.2.
51
Naruto v Slater, 888 F.3d 418 (9th Cir. 2018).
52
Ibid.

I led a team of patent attorneys in filing two patent applications in 2018 for two AI-generated
inventions—the “Artificial Inventor Project.”53
One invention was a beverage container based
on fractal geometry (like a snail shell) and one was a flashing light beacon that could attract
attention in an emergency. We filed those applications initially in the United Kingdom and the
European Patent Office (EPO) because those offices do not require an inventor to be initially
disclosed and because the offices subject the applications to early substantive review. Indeed,
the applications were reviewed and found by the UKIPO to be new, non-obvious, and useful at
a preliminary stage. That means if I had put my name, or just about anyone’s name, on those
patents as the inventor no one would have questioned me and we would likely have issued
patents by now. After substantive examination was complete, we updated the inventorship to
note that there was no human inventor and that an AI had functionally invented the inventions,
and we also filed the applications in 15 additional jurisdictions. Naming the AI as the inventor
was not done to give the AI ownership of a patent or to give it rights or credit. It was done to
ensure appropriate patent ownership, to be transparent, and to prevent someone from claiming
false credit. If I could license “DABUS,” which is the AI that generated our inventions, and
have it generate numerous inventions for which I could list myself as the inventor, that would
change what it means to be a human inventor. It would equate the efforts of people asking an
AI to solve a problem with someone who exhibited genuine ingenuity.
Of course, it has never been our position, or just about anyone’s position, that the AI should
own a patent. An AI could not legally own property because it is not a legal person, but more
importantly, an AI would not care about getting a patent and it could not properly exploit
a patent. Having an AI own a patent does not make a lot of sense on just about any level. I am
sure someone could figure out a way to make it work using blockchain, but there is really no
way in which the patent system does not more effectively achieve its goals by letting the AI’s
owner (or another default owner) patent the AI’s inventions. There are other parties that might
be appropriate default owners, such as programmers, users, or producers, but as long as there
is a default and clarity around subsistence, if multiple parties are involved they can contract to
their optimal outcome.
The cases thus have nothing to do with rights for AI. In the future, there could be some
reason to give an AI some sort of right. The idea seems less ridiculous when you realize
that artificial persons, mainly in the form of companies and governments, have all sorts of
rights, including civil rights (as Crouch explores). In fact, most patents are owned by artificial
persons. In terms of AI rights, for instance, we could change the law to allow a self-driving
car to have some limited form of legal personality so that it could hold an insurance policy to
directly compensate accident victims. But just because the law could change does not mean
that it should. I cannot see any way in which it makes more sense to give a self-driving Uber
an insurance policy than to simply make Uber liable for accidents caused by their self-driving
cars. If we did give a machine rights it would not be for the machine’s sake, but for the same
reason we give companies rights—we think that doing so makes things better for people.
Companies being able to own property and enter into contracts facilitates commerce and thus
benefits people.
In our patent cases, we argued the AI’s owner should own its patents as a default where
there are multiple claims of entitlement. That position is based on various common law rules of
53
See generally www.artificialinventor.com.

An introduction 17
property ownership. For example, if I own a 3D printer and I have it make a physical beverage
container, I own that beverage container. That outcome is based on the doctrine of accession,
which refers to someone owning a piece of property by virtue of owning some other piece of
property. For example, if I own a cow that has a calf, I own the calf. If I own a fruit tree that
bears fruit, I own the fruit. So, if I own an AI and it makes a patentable invention, I own that
invention as a trade secret (assuming it qualifies as such, as Sandeen and Aplin explore) if it is
not disclosed. If I choose to file for a patent on the invention, I should own that patent as well.
There are likely to be some interesting entitlement disputes in the future where many people
were involved in the making of an AI-generated invention. It is simple to say that if I own
a fruit tree I own its fruit, but the situation may be complicated. I might own the land on which
a tree is situated, but another person leases the land, another person supplies seeds, another
harvests the fruit, and so on. But the law has long dealt with competing claims of entitlement
by reference to underlying principles of property law. In our case, where only one party could
credibly claim entitlement, if DABUS’s owner gets to patent its output that creates all the right
incentives for the public to benefit.
To give a little background on how DABUS works,54
at its simplest, consider a system
of two neural networks. A network is composed of numerous nodes connected by the same
algorithm. The first network is trained on data, which alters the connection weights between
the nodes and which essentially stores the data. For instance, you could train the network by
exposing it to 100,000 car suspension designs. The system then generates noise by altering its
own connection weights, essentially corrupting the data it has been trained on, which generates
novel output—say, variations on the car suspension designs it was initially trained on. The
second network knows what data the first network was trained on, so it can tell you whether
what is coming out of the first network is new (and how different), and it can control the level
of noise in the first network. It can also be trained to model the first network’s output, so, for
instance, it can take a proposed car suspension design and evaluate it for fitness—how well it
performs across certain measures, such as weight. Set the system up properly and you can have
one network pumping out new designs at superhuman speeds, and another network evaluating
how well those will perform. At some point that should result in a better car suspension.
Two networks is the simple version, but in modern times a system like this can be composed
of hundreds or thousands of neural networks, each representing a concept such as warmth or
enjoyment. A person trains the AI in how simple concepts relate to each other—warm food
can result in enjoyment, for example. Later, in unsupervised operation, the machine combines
basic ideas into complex ideas and stops when a complex idea terminates in a particularly
salient concept. DABUS was not told to invent a flashing emergency light, but it was told to
be on the lookout for things that could prevent death. It combined the ideas for a new flashing
light mechanism with preferentially attracting attention with the need to attract attention in
an emergency, and essentially generated a patent claim that we built a specification around.
In our case, no one gave DABUS a specific problem to solve, it was not trained or built to
solve a specific problem, and DABUS identified the value of its output before it was seen by
a human being.
54
For a lot of background on how DABUS works, see Stephen L. Thaler, Vast Topological Learning
and Sentient AGI, 8(1) Journal of Artificial Intelligence and Consciousness 81 (2021).

In July 2021 we received our first patent in South Africa.55
DABUS is listed as the inventor
and the patent belongs to the AI’s owner. South Africa is a little unusual in that it does not
substantively examine patents.56
In other words, South Africa does not consider whether
inventions are new, non-obvious, and useful before issuing them. But the applications had
already been reviewed for substantive patentability by the UKIPO and EPO. South Africa does
do formalities examination, and in every jurisdiction that has so far denied the applications this
has been on a formality basis.
Not every jurisdiction has been as open-minded as South Africa. The applications are
pending in 16 jurisdictions but have so far been denied by courts and patent offices in several
of those, including in the United States, United Kingdom, EPO, and Australia, on the basis
that the applications did not properly designate a human inventor. All those denials are under
appeal, and in July 2021 Justice Beach in the Federal Court of Australia (FCA) issued an
extensive reason decision essentially holding that an AI could be a patent inventor as a matter
of law and that, at least in our case, the AI’s owner had the most compelling claim of enti-
tlement.57
That decision is currently being reviewed by a full panel of the FCA, from which
a further appeal to Australia’s High Court is available on a discretionary basis.
In the United States, the USPTO denied the applications and this denial was upheld by
a Justice Brinkema in the Eastern District of Virginia.58
An appeal is currently before the
Court of Appeals for the Federal Circuit. In the United States, the USPTO argues that the
Patent Act’s plain language refers to inventors as persons, and therefore (1) an AI cannot be
an inventor and (2) absent a traditional human inventor a patent cannot be obtained. I argue
that the Patent Act never defines an inventor as a natural person, and a broader interpretation
is required in light of the context and purpose of the Patent Act.
In September 2021 the United Kingdom Court of Appeal upheld a High Court decision
upholding the UKIPO’s denial of the applications.59
However, Justice Birss did hold in our
favor that there should be no prohibition on our receiving a patent for an AI-generated inven-
tion.60
Justice Arnold and Justice Liang did not concur. The United Kingdom Supreme Court
is currently deciding whether to accept an appeal from the Court of Appeal. By the time you
are reading this, of course, that entire summary is likely to be outdated.61
It may also be that legislative change is needed to allow patents on AI-generated inventions
in some jurisdictions. India recently completed a parliamentary consultation which held that the
law should be amended to explicitly provide copyright and patent protection for AI-generated
55
ZA2021/03242, https://iponline.cipc.co.za/Publications/PublishedJournals/E_Journal_July
%202021%20Part%202.pdf.
56
Ed Colon, DABUS: South Africa issues first-ever patent with AI inventor, www.managingip.com/
article/b1sx9mh1m35rd9/dabus-south-africa-issues-first-ever-patent-with-ai-inventor (last accessed
02/27/2022).
57
Thaler v Commissioner of Patents [2021] FCA 879.
58
Thaler v Iancu, et al. (No. 1:20-cv-00903).
59
Thaler v Comptroller General of Patents Trademarks and Designs [2021] EWCA Civ 1374.
60
Ibid at 22 (“The fact that the creator of the inventions in this case was a machine is no impediment
to patents being granted to this applicant”).
61
But should be updated at www
.artificialinventor
.com.

An introduction 19
works.62
After that, the President of South Korea announced that he believed AI-generated
inventions should be protected and that patent law needed to accommodate this.63
We filed these test cases for several reasons. The first was to generate some guidance for
industry about how to structure the use of AI in RD. There were a lot of assumptions about
how the law would handle AI-generated works, but little law on the subject or judicial analy-
sis. The second was to encourage a broader discussion about IP and AI, and how IP systems
should respond to technological advances. The third was to advance the normative position
that promoting patents on AI-generated inventions would benefit society. I suppose there was
also a hope that a case involving inventive AI would make patent attorneys marginally more
interesting than tax attorneys.
Again, AI-generated works are just one example of how AI will challenge long-standing IP
laws. Another way in which AI will impact patent law will be its effect on the test for whether
a patent is non-obvious (the focus of Shemtov and Gabison’s chapter).64
This test focuses on
the capabilities of an average researcher, but in many fields average researchers are already
augmented by AI. AI provides researchers with access to a superhuman amount of prior art
(existing knowledge), and it provides tools to help solve problems. For instance, a person
might require exceptional skill to recognize a pattern in a large data set, but with the help
of AI that activity might be trivial. Thus, it may already be that the “skilled person” should
be a skilled person using AI. That will raise the bar to getting patents, because more will be
obvious to a skilled person using AI. In fact, widespread augmentation with AI may make it
very difficult to get a patent. Consider the best-selling biological drugs, which are all mono-
clonal antibodies. These are like antibodies made by your body, but the drug form is composed
of copies of the same antibody which can be manufactured using a variety of biological pro-
cesses. These antibodies can treat cancer or pathogens such as COVID-19, so they are a very
important type of treatment.
One of the interesting things about antibodies from an IP and AI perspective is that there
are a finite number of antibodies that can exist. Antibodies are combinations of amino acids
with a particular structure, and there are only so many amino acids and only so many ways to
combine them to make an antibody. To be certain, there are at least trillions of possible anti-
bodies, so enough options to keep human researchers busy for the foreseeable future. But there
are not so many that an AI could not sequence every possible antibody. Moreover, AI is getting
better at modeling antibody activity, so the AI could both generate every possible antibody
and provide useful information about the sorts of biological activity each antibody will have in
the body. If an AI did that, and if that information was published online, it might prevent any
person from ever patenting a new antibody as a composition of matter. That claim and subject
matter usually makes up the most important patent in a portfolio for a new biological drug.65
Many computer scientists believe we will eventually develop Artificial General Intelligence,
a future AI that will still not think like a person but that will be able to do any intellectual task
62
Varsha Jhavar, Parliamentary Standing Committee’s recommendations concerning AI and
IP: a little late or way too early? https://spicyip.com/2021/08/parliamentary-standing-committees
-recommendations-concerning-ai-and-ip-a-little-late-or-way-too-early.html (last accessed 02/27/2022).
63
www.facebook.com/worker21c/posts/4223612477726208 (last accessed 02/27/2022).
64
Ryan Abbott, Everything is Obvious, 66 UCLA. L. Rev. 2 (2019).
65
See Ryan Abbott, Hal the Inventor: Big Data and Its Use by Artificial Intelligence, in Big Data
is Not a Monolith (Cassidy R. Sugimoto, Hamid Ekbia and Michael Mattioli eds, 2016), exploring
similar hypotheticals.

a person could do. The first task for that AI would be to improve its own programming and
capabilities, and theoretically this may lead to an AI so sophisticated that it will solve all the
problems we know we have and even those we do not know we have.66
There are more exciting
things to think about with superintelligent AI than patent law, but among other things, it would
make it virtually impossible to get a patent because everything would be obvious to a person
using a superintelligent AI (or at some point maybe we jettison the person and just have
a superintelligent AI standard). That should be fine, though, because once we have superintel-
ligent AI there will no longer be a need to incentivize innovation, which will be nearly costless.
There will also be little need to encourage disclosure of AI output because trade secrets will
be quickly independently replicated by other superintelligent AI. To the extent that we need
a system to encourage commercialization of products such as medicines where most of the
RD cost is after an initial invention, we have other means of encouraging that activity, such
as a period of market exclusivity associated with regulatory approvals.
4. CONCLUDING THOUGHTS
Even for readers who do not work directly with AI and IP, the topic still offers valuable
insights into IP law more generally. For example, we think that patents and copyright both
protect economic and moral rights, so copyright protects both an author’s right to be acknowl-
edged and a publisher’s ability to make money. If AI does not have rights and a human author
is no longer in the picture, is there still a case for copyright if we are no longer concerned with
protecting authors in this context? Do we still want to apply copyright the same way? Do we
need human-centric laws? Do the laws that currently apply to people also change once AI is in
the picture? People talk about different things when they talk about AI-generated inventions,
but they also talk about different things when they talk about access to medicines, or the patent
system generally, or any number of traditional concepts in IP law.
My own view is that treating AI and human behavior differently tends to result in problem-
atic outcomes for people. In patents, if the system is designed to generate innovation, then we
should leave it to the market to decide whether people or AI can do a better job of inventing—
because if you tell GSK or Johnson Johnson that they cannot use an AI to make a drug, at
least not if they want a patent on that drug, then it tells industry it has to use people. That will
be a real problem if it turns out that machines can do a better job than people at certain sorts
of important activities.
Similar dynamics occur in other areas of the law. Businesses using AI to automate work
are subject to different tax consequences versus having people do that same work. Cars driven
by people versus AI generate different liability regimes. In The Reasonable Robot: Artificial
Intelligence and the Law I look in depth at a few areas, including in IP law, where AI and
people functionally do the same thing but different laws apply and this ends up having socially
harmful consequences.67
I explain why we would all be better off if the law tended not to dis-
criminate between human and AI behavior, a principle I refer to as AI legal neutrality.
AI and board games may offer a useful roadmap for IP and AI. AlphaGo beating the world’s
best Go player was a big deal, even if it did not get as much attention (in the United States)
66
Ryan Abbott, The Reasonable Robot (2020) at 24.
67
See generally ibid.

An introduction 21
as DeepBlue beating Gary Kasparov at chess in 1997.68
Kasparov subsequently realized that
while an AI might always be able to outperform a person, an AI playing with a person could
outperform an AI, because people and machines have complementary capabilities. He then
won the first “centaur chess” tournament involving a person and an AI playing together. This
points to a near and medium-term future in which we are going to get optimal outcomes by
combining AI and human behavior. But in the long term, eventually AI improved to the point
where playing with a human chess grandmaster just slowed it down.
It would be a phenomenal outcome if, the next time a deadly pathogen such as COVID-19
emerges, AI has become so proficient at developing vaccines that having human vaccinologists
in-the-loop just makes the development process less effective. How soon that comes to pass,
and perhaps whether it comes to pass at all, will depend in part on our IP rules. Hopefully, this
Handbook will help stakeholders to better understand how to think about IP and AI, and thus
contribute to rules that result in better outcomes for people.
68
Deep Blue defeats Garry Kasparov in chess match, www.history.com/this-day-in-history/deep
-blue-defeats-garry-kasparov-in-chess-match (last accessed 02/27/2022).

22
2. The human cause
Daniel J. Gervais1
1. INTRODUCTION
Oxford University Professor Nick Bostrom’s book Superintelligence: Paths, Dangers,
Strategies describes an advanced future form of artificial intelligence (AI) which, he says, will
be the last invention made by humans.2
Though this “superintelligence” belongs to the world
of science fiction, there is no doubt that there are “challenges posed by highly intelligent (ro)
bots participating with humans in the commerce of daily life.”3
Nowhere are those challenges
more manifest than when those machines are able to perform tasks that until recently only
humans could, namely tasks anchored in our higher mental faculties, including human crea-
tivity and our ability to develop innovative technologies. The area of law that is most likely to
feel the impact of this emergence of a second intelligent and potentially creative and inventive
“species” is intellectual property.4
The ability of AI to produce literary and artistic works and inventions matters on another
level.5
By now, we are used to letting robots perform much of the physical labor previously
done by humans.6
With less of that kind of work, our capacity to create art, music, literature,
conversation, architecture, food, and more “is likely to be more needed than ever.”7
Yet,
as things stand now, we are “devoting huge scientific and technical resources to creating
ever-more-capable AI systems, with very little thought devoted to what happens if we suc-
ceed.”8
If machines become creators and inventors in our stead, will we be able to, as Keynes
aptly put it, “keep alive, and cultivate into a fuller perfection, the art of life itself”?9
It is at
least worth pondering. To state but one reason, changes in cultural productions and trends
1
This chapter is based in part on the author’s 2020 Sir Hugh Laddie Lecture. The author is grateful
to the editor both for his comments and for inviting this contribution.
2
There is no universally agreed upon definition of artificial intelligence. This chapter adopts
the definition used by the European Commission: “Artificial intelligence (AI) refers to systems that
display intelligent behaviour by analysing their environment and taking actions—with some degree
of autonomy—to achieve specific goals.” European Commission, Communication on “Artificial
Intelligence Europe,” COM(2018) 237 final (April 25, 2018), online https://
bit
.ly/
2HFft4J.
3
Wendell Wallach Collin Allen, Moral Machines: Teaching Robots Right from
Wrong 189 (Oxford University Press, 2009).
4
Admittedly, I use the term “species” in a non-technical way in this context.
5
In this chapter, human creativity and inventiveness will be referred to as “natural,” while machine
outputs (or productions as the term is used hereinafter) will be referred to as “artificial,” following the
same logic as the terms “natural language” and “artificial intelligence.”
6
On robots replacing human workers, see Bill O’Leary, My Robot And Me, 68:6 Electrical
Apparatus 18 (2015).
7
Stuart Russell, Human Compatible: Artificial Intelligence and the Problem of Control
122 (Viking Press, 2019).
8
See Wallach Allen, supra note 3, at 151.
9
John Maynard Keynes, Essays on Persuasion (Norton, 1963) at 331.

The human cause 23
both lead and reflect societal changes, which in turn lead to political and, ultimately, legal
changes. Literature in all forms, fine arts, and music are among the most important vehicles to
both mirror and propagate changes throughout society. If those cultural vehicles are made of
art, books, and lyrics created by AI machines, then those machines will control at least a part
of cultural, societal, and political change. Think of it as self-driving culture—and it will be
a U-turn as far as human evolution is concerned.
This chapter cannot predict whether human authors and inventors will survive as a signif-
icant source of cultural production and technological innovation in the medium to long term.
On a shorter time horizon, however, a fair question to ask is whether intellectual property
should prioritize human (or natural) creativity and inventiveness, or else treat machine and
human productions on the same footing and accelerate the replacement of human creators and
inventors. As used in this chapter, giving “priority” would mean granting rights only in facially
copyrightable or patentable productions that have a human cause, which can be provisionally
defined for now as a sufficient link between one or more humans and the potentially copy-
rightable or patentable output. 10
This explains the double-entendre in the title of the chapter:
while the chapter discusses whether intellectual property protection should only attach to
creations and inventions that have an identifiable and sufficient human cause, it also implies
that the future of humans (their “cause”) is involved. As the chapter explains below, the notion
of cause used here is similar to proximate cause, not simple (or “but for”) cause.
One can use the US Constitution as a useful backdrop for the analysis. It is unique among
constitutional documents in that it not only gives one level of government (federal) the power
to make laws in the area of copyright and patents, it actually states what the purpose of those
laws is, or should be: to “Promote the Progress of Science and Useful Arts.”11
What if the
novelists, songwriters, journalists, and inventors of tomorrow were machines? What if instead
of human actors, we had CGI “actors” looking like famous (dead) human actors or, once those
have been mostly forgotten, CGI “actors” with AI-designed faces and bodies meant to appeal
to the largest number of human viewers? What if the role of humans—other than for the few
who would profit from this situation—was relegated to reading novels and news, listening to
music, watching audiovisual content, and buying new products produced by machines? Were
the Founding Fathers thinking of “progress” in those terms, that new art and science would
be produced for its own sake, not for human progress? To answer, let us turn the knob all the
way: if all humans died would the US Constitution’s direction still be followed provided art
and science continued to be produced—both by machines, and for machines? Thus to say that
law should be technologically neutral is the same as saying it can be human neutral.
We need not turn the knob all the way to find the path to a policy prescription. If machines
can produce new drugs or literary and artistic works cheaper and faster than human creators,
it is highly likely that industry will favor them over their human counterparts. In the copyright
sphere, delegating to machines the task of helping us understand and interpret our world has
profound consequences. It is through this interpretation that humans can become true agents
in the world and ultimately change it. Delegating this very task to machines is thus pregnant
with implications for the future, for it changes its arc. It will not be complete obliteration, of
course. There will always be humans who write, pick up a paintbrush, or try to make a movie
10
As used in this chapter, the term “production” encompasses (a) creations that may be protected by
copyright, (b) designs, and (c) inventions that may be protectable by patents.
11
U.S. Const, Art. 1, §8, cl.8. This chapter knows of no other constitutional document that does so.

or sculpture, but if most of what we are given to read, watch, or listen to comes from machines,
much will be lost. If copyright protection is granted on productions without a human cause,
and assuming that the cost of machine productions will be lower (and machines will not ask for
ongoing royalty payments or have reversion rights), then market forces will inescapably push
for a replacement of human authors whenever it is commercially feasible.
In science, if machines can both do basic science and develop new technologies, as exam-
ples in this volume demonstrate, private labs will hire fewer science PhDs. Science depart-
ments in universities will atrophy as the demand for human-made science contracts. The arc of
the quest for a deeper and better understanding of the natural world that has animated humans
arguably since the invention of fire will also be bent. In sum, our highest and noblest ideals
will be delegated, at least in part, to machines. This should perhaps give us pause.
While, as other chapters in this book illustrate, we can debate what exactly constitutes ‘pro-
gress’ (especially, one might add, from a postmodernist perspective), this chapter’s normative
sextant is that the term “progress” must mean human progress, in the manner of philosophers
through the ages.12
For example, Plato defended in his Laws the idea that the legal system is
a way to support human progress.13
Even Aristotle might have agreed, as human flourishing
is a core notion of his Nicomachean Ethics. According to this view, “the invention of man is
infinitely better contrived to advance the good and happiness of mankind, than any Utopian
system that ever has been produced, by the warmest imagination.”14
Of the imagination of
a machine?
Let us begin by making three general points before diving deeper into the analysis. The first
is that certain AI machines can be programmed to learn to mimic human mental processes.
Part of the ongoing research in AI is precisely to make machines more like humans instead
of a new, complementary type of intelligence. As a result, there is little doubt that machines
already generate productions that are often indistinguishable from human creations and inven-
tions. From that perspective, giving rights on productions that look like copyright works or
patentable inventions but made by machines strikes the author of this chapter as a new Turing
test.15
Normatively, it would amount to rewarding the owner or user of a machine that can
“pass itself off” as human. Surely that anthropomorphizing illusion cannot be a solid norma-
tive foundation to obtain IP protection.16
12
See generally Daniel Gervais, The Law of Human Progress (deLex, 2019).
13
Plato, Laws, VII:680a-682d.
14
Henry H. Kames, Essays On the Principles of Morality and Natural Religion, 86. See
also J.J. Chambliss, Human Development in Plato and Rousseau 13:2 The Journal of Educational
Thought 96, 98 (1979).
15
That may remind the reader of the “Turing test,” a set of questions asked via teletype on any
subject whatsoever. Unbeknown to the questioner, some were answered by a human and others by
a machine. Both the human being and the machine attempted to convince the questioner that it or she was
the human and the other was not. See Lawrence B. Solum, Legal Personhood for Artificial Intelligences,
70 N.C. L. Rev. 1231, 1236 (1992).
16
See Selmer Bringgjord David A. Ferrucci, Artificial Intelligence and Literary
Creativity: Inside the Mind of Brutus xxvi (Lawrence Erlbaum Associates, 2000). Passing off is
a tort notion borrowed from trademark law defined as “when a producer misrepresents his or her own
goods or services as those of another producer.” Laura Gasaway, Origin of Goods in Trademark Law
Does Not Mean Creator; Copyright Corner, Special Libr. Ass'n Info. Outlook, Nov. 1, 2003, at 7. See
also 15 U.S.C. §1125(a) (“palming off”); and Dastar Corp. v Twentieth Cent. Fox Film Corp., 539 U.S.
23, 28 n.1 (2003).

The human cause 25
Second, humans and machines are working ever more closely together. This rapprochement
will continue.17
Humans depend on machines to perform many creative and inventive tasks,
and indeed, machines have already changed how humans perform those tasks. In some cases,
machines can help us achieve our aims better and faster. It is not always so. To take a simple
example of a change in cognitive processes with a more ambivalent valence, people who
started driving a car before GPS was omnipresent can still drive in cities where they drove
before the GPS without assistance from that technology, but are much less able to do so
elsewhere without GPS.18
A key question to answer in this context is: What happens over the
medium to long term as we outsource the creative or inventive work of humans to machines?
Humans, as a species, may lose on two fronts: diminished human expression, and reduced
financial flows to human creators and inventors, who would no longer have the incentive,
time, or financial ability to learn and develop their craft. To say that creativity is necessarily
human, that it is fundamentally connected with humanness, is not “to impose a kind of chau-
vinism that privileges human-produced artifacts over those that are machine-made. Rather, it
is to say that human communication is the very point of authorship as a social practice; indeed,
as a condition of life.”19
The third and final general point is that there are proposals to short-circuit that entire discus-
sion by giving “person” status (which is not the same as “human” status, of course) to some AI
machines.20
The root of the first word in the term “artificial intelligence,” namely “artificial,”
is “artifice,” the definition of which is “an ingenious device or expedient.”21
This is apt because
the proposal to give some AI machines legal personality may be just that: an expedient that
circumvents the two underlying normative issues, namely whether such machines should be
persons and should get IP rights.22
2. BRIEF OVERVIEW OF THE STATE OF PLAY IN AI
A significant portion of the literature exploring the interface between AI and the law adopts
the distinction between narrow (or weak) and general (or strong) AI.23
The distinction can be
17
One should not avoid drawing the line between human and machine, which will not always be easy
as cyborgization increases, but the legal system must be able to draw that line.
18
Javadi, Amir-Homayoun et al., Hippocampal and Prefrontal Processing of Network Topology
to Simulate The Future, 8 Nature Comm. 14652 (2017), online: https://www.nature.com/articles/
ncomms14652#citeas.
19
Carys Craig and Ian Kerr, The Death of the AI Author, 52 Ottawa L Rev 31–86 (2021).
20
See, e.g., Shawn Bayern, Of Bitcoins, Independently Wealthy Software, and the Zero-Member
LLC, 108 Nw. U.L. Rev. 1485, 1497 (2014).
21
“Artifice,” Merriam-Webster Dictionary, www.merriam-webster.com/dictionary/.
22
See David J. Calverley, Imagining a Non-Biological Machine as a Legal Person, 22 AI Soc.
523 (2008). This is in line with a United Nations report which noted that it would be “highly counterin-
tuitive to call [AI systems] ‘persons’ as long as they do not possess some additional qualities typically
associated with human persons, such as freedom of will, intentionality, self-consciousness, moral agency
or a sense of personal identity”). UNESCO, World Comm. on the Ethics of Scientific Knowledge and
Technology (COMEST), Rep. of Com. on Robotics Ethics, U.N. Doc. SHS/COMEST-10/17/2 REV., at
46 (Sept. 14, 2017).
23
See Michael Guihot, Anne F. Matthew, Nicolas P. Suzor, Nudging Robots: Innovative Solutions
to Regulate Artificial Intelligence, 20 Vand. J. Ent. Tech. L. 385, 393 (2017). See also Shannon

traced back to Ray Kurzweil’s seminal The Singularity Is Near.24
In this traditional categori-
zation, narrow or weak AI “is goal-oriented, designed to perform singular tasks—i.e. facial
recognition, speech recognition/voice assistants, driving a car, or searching the Internet—and
is very intelligent at completing the specific task it is programmed to do.”25
It operates within
a well-defined “activity-context.”26
In contrast, Artificial General Intelligence (AGI), or strong
AI, “is the concept of a machine with general intelligence that mimics human intelligence and/
or behaviours, with the ability to learn and apply its intelligence to solve any problem. AGI
can think, understand, and act in a way that is indistinguishable from that of a human in any
given situation.”27
A number of scholars go a step further. Nick Bostrom, who was mentioned
in the opening paragraph, discusses the risks to humans of developing machines with a higher
level still, which he dubbed “superintelligence,” that is, “an intellect that is much smarter than
the best human brains in practically every field, including scientific creativity, general wisdom
and social skills.”28
This Artificial Super Intelligence (ASI) “is the hypothetical AI that doesn’t
just mimic or understand human intelligence and behaviour; ASI is where machines become
self-aware and surpass the capacity of human intelligence and ability.”29
This chapter retains this traditional categorization but sticks to the first two categories
(narrow and strong), as it sees the third (ASI) category as belonging to the world of sci-fi—at
least for now. Machines can beat the best human masters at chess, Go, the gameshow Jeopardy
and much more—even at an incomplete information game like poker.30
Though undoubtedly
very impressive, those achievements “are much simpler than the real world: they are fully
Vallor George A. Bekey, Artificial Intelligence and the Ethics of Self-Learning Robots, in Robot
Ethics 2.0: From Autonomous Cars To Artificial Intelligence 339–40 (Patrick Lin, Keith Abney,
Ryan Jenkins eds, 2017); and Peter Stone et al., Artificial Intelligence And Life In 2030: Report of the
2015 Study Panel 6–9 (2016), https://ai100.stanford.edu/sites/default/files/ai_100_report_0831fnl.pdf.
24
Ray Kurzweil, The Singularity Is Near: When Humans Transcend Biology 206, 222 (Rick
Kot ed., 2005).
25
Serena Reece, What Are The 3 Types Of AI? A Guide To Narrow, General, And Super Artificial
Intelligence, CodeBots, Jan. 31, 2020, online: https://codebots.com/artificial-intelligence/the-3-types-of
-ai-is-the-third-even-possible.
26
Stephen Russell, Ira S. Moskowitz, Adrienne Raglin, Autonomy and Artificial Intelligence:
a Threat or Savior?, in Autonomy and Artificial Intelligence: A Threat or Savior?, 71, 73 (W.F.
Lawless et al., eds).
27
Reece, supra note 25.
28
Nick Bostrom, How Long Before Superintelligence?, 5 Linguistic Phil. Investigations 11, 11
(2006). Updated version (2008), online at www
.nickbostrom
.com/
superintelligence
.html.
29
Reece, supra note 25.
30
See Cade Metz, In Two Moves, AlphaGo and Lee Sedol Redefined the Future, Wired, Mar. 16,
2016, online: www.wired.com/2016/03/two-moves-alphago-lee-sedol-redefined-future/. The “former”
champion eventually retired as a result. See James Vincent, Former Go Champion Beaten By DeepMind
Retires After Declaring AI Invincible, The Verge (Nov. 27, 2019). The difference in the levels of
complexity is not the only one between chess and Go. Chess is tactical while Go is best described as
strategic. It requires a different kind of “thinking.” Cade Metz, In a Huge Breakthrough, Google’s AI
Beats a Top Player at the Game of Go, Wired, Jan. 27, 2016, online: www.wired.com/2016/01/in
-a-huge-breakthrough-googles-ai-beats-a-top-player-at-the-game-of-go/. On IBM’s DeepBlue beating
the world chess grandmaster Gary Kasparov, see IBM’s 100 Icons of Progress: Deep Blue, www.ibm
.com/ibm/history/ibm100/us/en/icons/deepblue/ [https://perma.cc/7SG3-UYST]. On poker, see Tracey
Lien, Artificial Intelligence Has Mastered Board Games; What’s The Next Test? Seattle Times
(Mar. 20, 2016), online www
.seattletimes. com/business/technology/artificial-intelligence-has-mastere
d-board-games-whats-the-next-test/.

The human cause 27
observable, they involve short time horizons, and they have relatively small state spaces and
simple, predictable rules.”31
This raises a related question, namely whether people will still have an incentive to play
games such as chess or Go professionally with the knowledge that an AI machine can beat
the best humans, even with a few computer chips tied behind its back. I believe the answer is
yes. I start from the premise that humans have always played and always will.32
If you accept
that premise, then if someone plays, say, chess, every day for years, they will become good at
it. They can then enter the human “rankings” (for example, grandmaster) system. Now, why
would other players want to see those people play even if they know they cannot fight the
machine and hope to win? For two main reasons. First, because humans are more likely to
learn from watching other humans and not machines, whose “thinking” may not be the same
as those of humans and who may not be able to explain their thinking to begin with.33
Second,
I posit that humans like to watch other humans “struggle.”34
To simplify to the extreme, this
is why we watch sports but also, say, a “strongman” pulling a ton of bricks with ropes—
something any half-decent pickup truck can do on a couple of cylinders. For similar reasons,
I am much more worried about machines replacing songwriters and composers than about
machines replacing live performers. People will likely want to see human artists/performers. It
is thus essential in debates about the future of copyright to distinguish authors from perform-
ers, a well-understood and fundamental distinction in the law of copyright and related rights.35
In the field of potentially patentable productions, AI is now routinely used to accelerate
and reduce the costs of pharmaceutical research, performing in silico research.36
AI machines
can find hidden patterns within large datasets and automate many predictions.37
Outputs from
AI in pharmaceutical research include disease diagnosis and prediction of drug efficacy38
and support for drug design.39
AI machines can choose which molecules possess suitable
characteristics to address biological targets of interest.40
AI can identify the optimal chemical
structures to reduce toxicity and satisfy metabolic requirements, both of which can be costly
and data-intensive processes.41
They can improve the area of personalized medicine based on
31
Russell, supra note 7, at 56. Emphasis added.
32
See generally Juho Hamari Lauri Keronen, Why Do People Play Games? A Meta-Analysis, 37:3
Int’l J. Inf. Manag’t 125 (2017).
33
The idea that humans and machines can both perform a function that can be described as “think-
ing” but do so differently is not new. See e.g., Philip N. Johnson-Laird., Human and Machine
Thinking (Lawrence Erlbaum Associates, 1993).
34
See Michael Safi et al, How Magnus Carlsen Won Chess Back From The Machines, The
Guardian (Dec. 12, 2021), available at https://
bit
.ly/
33n4Mlu.
35
See Daniel Gervais, Related Rights in United States Law, 65 J. Copyright Soc’y U.S.A. 371–93
(2018).
36
See Nic Fleming, How Artificial Intelligence Is Changing Drug Discovery, 557 Nature 55 (2018).
37
See generally Ajay K. Agrawal, Joshua S. Gans, Avi Goldfarb, Prediction Machines: The
Simple Economics Of Artificial Intelligence (Harvard Business Press, 2018).
38
See Gregor Guncar et al., An Application Of Machine Learning To Hæmatological Diagnosis, 8
Scientific Reports 411 (2018), online: https://
bit
.ly/
3qrclyS.
39
See Hongmin Chen et al., The Rise Of Deep Learning In Drug Discovery, 23 Drug Discovery
Today 1241–50 (2018).
40
See ibid.
41
See ibid.

genetic markers.42
That potential of AI to identify novel drugs that human researchers alone
cannot detect has attracted investment from both start-ups and established pharmaceutical
companies.43
Also noteworthy, in what was perhaps a publicity stunt, Google announced that
its AI machines can both make new inventions and apply for patents.44
In the fields of design and literary and artistic works, AI machines have composed poly-
phonic baroque music bearing the “style” of Johann Sebastian Bach.45
“Robot reporters” now
routinely write news bulletins and sports reports, a process called “automated journalism.”46
AI
systems write poems that many people believe were written by a human author.47
AI machines
draft and analyze contracts.48
A machine named e-David produces paintings using a complex
visual optimization algorithm that “takes pictures with its camera and draws original paintings
from these photographs.”49
AI machines can write scenes of animation movies and improve
the design of objects or processes, thus generating productions that facially qualify as subject
matter for copyright or design patent protection.50
Let us now briefly see how AI does it.
42
See Kit-Kay Mak Mallikarjuna Rao Pichika, Artificial Intelligence in Drug Development:
Present Status And Future Prospects, 24:3 Drug Discovery Today 773 (2019).
43
See Lou Bowen Lynn Wu, Artificial Intelligence And Drug Innovation: A Large Scale
Examination Of The Pharmaceutical Industry 2 (2020), online: https://papers.ssrn.com/sol3/papers.cfm
?abstract_id=3524985.
44
See Rose Hughes, Deepmind: First Major AI Patent Filings Revealed, IPKat, Jun. 7, 2018, online
http://ipkitten.blogspot.com/2018/06/deepmind-first-major-ai-patent-filings.html. The reverse use of
AI is true, namely to defeat patent applications, based on obviousness (to an AI expert) or novelty, by
massive preemptive public disclosure of novel subject matter together with its utility. On the former,
see Ryan Abbott, Everything Is Obvious, 66 UCLA L. Rev. 2, 40 (2019). On the latter issue, see Daniel
Gervais, Exploring the Interfaces Between Big Data and Intellectual Property Law, 10:3 J. Intell.
Prop., Inf. Tech. E-Comm. L. (2019), online: https://www.jipitec.eu/issues/jipitec-10-1-2019/4875.
45
See Gaëtan Hadjeres François Pachet, Deepbach: A Steerable Model for Bach Chorales
Generation (Dec. 3, 2016) at 1, online: https://
arxiv
.org/
pdf/
1612
.01010v1
.pdf.
46
See Corinna Underwood, Automated Journalism—AI Applications at New York Times, Reuters,
and Other Media Giants, eMerj (Jun. 22, 2017, updated Nov. 29, 2018), online: https://
bit
.ly/
2Q84BTV.
See also Lucia Moses, The Washington Post’s Robot Reporter Has Published 850 Articles in the Past
Year, DigidayUK, Sept. 14, 2017, online: https://
bit
.ly/
2xmkQSI.
47
See Samuel Gibbs, Google AI Project Writes Poetry Which Could Make Vogon Proud, The
Guardian (May 17, 2016).
48
See Kathryn D. Betts Kyle R. Jaep, The Dawn of Fully Automated Contract Drafting: Machine
Learning Breathes New Life into A Decades-Old Promise, 15 Duke L. Tech. Rev. 216 (2017).
49
See Shlomit Yanisky-Ravid, Generating Rembrandt: Artificial Intelligence, Copyright, and
Accountability In The 3a Era—The Human-Like Authors Are Already Here—A New Model [2017] Mich.
St. L. Rev. 659, 662.
50
On copyright, see Jane C. Ginsburg and Luke Ali Budiardjo, Authors and Machines, 34 Berk.
Tech. L. J. 343 (2019). The United States Court of Appeals for the Federal Circuit noted that processes
“that automate tasks that humans are capable of performing are patent-eligible if properly claimed.”
McRO, Inc. v Bandai Namco Games Am. Inc., 837 F.3d 1299, 1313 (Fed. Cir. 2016). For a discussion,
see Ben Hattenbach Gavin Snyder, Rethinking the Mental Steps Doctrine and Other Barriers to
Patentability of Artificial Intelligence, 19 Colum. Sci. Tech. L. Rev. 313, 317–18 (2018); and Mizuki
Hashiguchi, The Global Artificial Intelligence Revolution Challenges Patent Eligibility Laws, 13 J. Bus.
Tech. L. 1, 13 (2017).

The human cause 29
3. HOW AI WORKS
This section reviews a few basic notions of AI that will be important as we attempt to draw
conclusions later on.
The deployment of AI can be separated into steps. First, AI code is written. This code,
as the technology stands now, is generally the work of human programmers—though that
is changing—and it can be split into pieces, such as a generic AI platform and specific apps
developed for a precise purpose.51
The code is mostly used to empower the next step, a process
known as machine-learning, which today is “the dominant AI technology.”52
Machine-learning
can be supervised (by humans) or not. “Unsupervised” in this context means that the system is
“trained on a dataset without explicit instructions or labelled data.”53
Situated between super-
vised and unsupervised learning, reinforcement learning is a third mode of machine-learning,
in which humans verify what the machine has learned on its own and hopefully correct mis-
takes, often using sampling techniques.54
Machine-learning in all three modes is used both to discern and operationalize patterns in
data.55
It uses a set of “computational methods using experience to improve [its] performance
or to make accurate predictions.”56
Using machine-learning, an AI system can “automatically
generate heuristics” and make autonomous determinations of various kinds.57
It can adjust
its “behavior to enhance [its] performance on some task through experience.”58
A machine
can, for example, be shown pictures of cats and dogs and then learn the features of each so
that it can distinguish cats and dogs it has never “seen” before.59
The quality of the learning
process is obviously dependent on the quality of the training data, as some well-documented
disastrous examples have brought to light.60
This is a problem for uses of AI in a legal context,
51
On AI machines writing their own code, see Khari Johnson, AI Could Soon Write Code Based
on Ordinary Language, Wired (May 26, 2021), online: www.wired.com/story/ai-write-code-ordinary
-language/(accessed May 31, 2021).
52
UK Information Commissioner’s Office and Alan Turing Institute, Explaining How Decisions Are
Made with AI, 7 (May 20, 2020), online: https://
bit
.ly/
2zs68gi. See also Roberto Iriondo, Differences
Between AI and Machine Learning and Why it Matters, Data Driven Investor, (Oct. 15, 2018), https://
medium.com/datadriveninvestor/differences-between-ai-and-machine-learning-and-why-it-matters
-1255b182fc6.
53
See UK Information Commissioner’s, ibid.
54
See Leslie Pack Kaelbling, Michael L. Littman, Andrew W. Moore, Reinforcement Learning:
A Survey, 4 J. Artificial Intelligence Res. 237 (1996).
55
Michael Veale, Governing Machine Learning that Matters, PhD dissertation, 33 (2019), online:
https://discovery.ucl.ac.uk/id/eprint/10078626/1/thesis_final_corrected_mveale.pdf.
56
Mehryar Mohri, Afshin Rostamizadeh Ameet Talwalkar, Foundations of Machine
Learning, 2d ed. 1 (MIT Press, 2018).
57
Wolfgang Hertel, Introduction to Artificial Intelligence 102 (Springer, 2011). AI
programmers use several different algorithmic techniques, depending (usually) on the task at hand. For
a detailed overview, see Explaining How Decisions Are Made with AI, above note 52, Annex 2.
58
Harry Surden, Machine Learning and the Law, 89 Wash. L. Rev. 87, 89 (2014).
59
See Amanda Levendowski, How Copyright Law Can Fix Artificial Intelligence’s Implicit Bias
Problem, 93 Wash. L. Rev. 579, 592 (2018).
60
For example, when Google’s AI created a link between images of African Americans and
gorillas. See James Vincent, Google “Fixed” Its Racist Algorithm by Removing Gorillas from Its
Image-Labeling Tech, The Verge (Jan. 12, 2018), online: www.theverge.com/2018/1/12/16882408/
google
-racist
-gorillas
-photo
-recognition
-algorithm
-ai; or when a new Microsoft AI chatbot quickly

for example when AI machines used in bail and sentencing decision-making reflect racial
or socio-economic biases due to the poor quality of the training data that was selected.61
Put bluntly, in some cases “[m]achine learning is a ‘garbage in-garbage out’ proposition.”62
As the many examples of AI achievements in the previous section demonstrate, however,
machine-learning can also be both quite powerful and productive. In sum, the quality and size
of the data “are crucial to the success of the predictions made by the [AI] learner.”63
The machine-learning function can take the form of “deep learning,” a subset of
machine-learning using a layered structure of algorithms allowing the machine to learn and
make predictions and decisions on its own.64
Deep learning has been called “the true challenge
to artiﬁcial intelligence,” namely solving the tasks that are easy for people to perform but
hard for people to describe formally—problems that we solve intuitively, that feel automatic,
like recognizing spoken words or faces in images.’65
With deep learning, one could say—
acknowledging that metaphors are intellectual shortcuts—that the computer has its own,
autonomous brain.66
Importantly, deep learning is automated and often (if not almost always)
removed from direct human input or control.67
There are various ways to make AI systems learn and perform better. One of them is the
development of General Adversarial Networks (GANs), a technological path likely to grow
the affordances of AI systems both qualitatively and quantitatively.68
“GANs’ potential is
huge, because they can learn to mimic any distribution of data. That is, GANs can be taught
to create worlds eerily similar to our own in any domain: images, music, speech, prose.”69
turned racist by “learning” on social media. See James Vincent, Twitter Taught Microsoft’s AI Chatbot to
Be a Racist Asshole in Less Than a Day, The Verge (Mar. 24, 2016), online: www
.theverge
.com/
2016/
3/24/11297050/tay-microsoft-chatbot-racist.
61
Many claims were made along those lines in various press and other sources, the truth and the
scope of which this chapter cannot independently verify. See, e.g., Cade Metz, We Teach A.I. Systems
Everything, Including Our Biases, N.Y. Times (Nov. 21, 2019) ; Kari Paul, Healthcare Algorithm Used
Across America Has Dramatic Racial Biases, The Guardian (Oct. 25, 2019); and Ed Pilkington, Digital
Dystopia: How Algorithms Punish the Poor, The Guardian,(Oct. 14, 2019).
62
Emily Berman, A Government of Laws and Not of Machines, 98 B.U. L. Rev. 1277, 1302 (2018).
63
Mohri, Rostamizadeh, Talwalkar, supra note 56, 1.
64
See Robert D. Hof, Deep Learning: With Massive Amounts of Computational Power, Machines
Can Now Recognize Objects and Translate Speech in Real Time. Artificial Intelligence Is Finally Getting
Smart, MIT Tech. Rev., www.technologyreview.com/s/513696/deep-learning/
65
Ian Goodfellow, Yoshua Bengio, Aaron Courville, Deep Learning 1 (MIT Press, 2016).
66
See Brett Grossfeld, A Simple Way to Understand Machine Learning vs Deep Learning, Zendesk
(Jul. 18, 2017), online www.zendesk.com/blog/machine-learning-and-deep-learning/. See also Claudio
Masolo, Supervised, Unsupervised and Deep Learning, Towards Data Science (May 7, 2017) online:
https://bit.ly/2BydnE8.
67
This has now gone mainstream. See William Vorhies, Automated Deep Learning—So Simple
Anyone Can Do It, Data Science Central (April 10, 2018), online: www
.datasciencecentral
.com/
profiles/blogs/automated-deep-learning-so-simple-anyone-can-do-it.
68
Indeed, Yann LeCun, FaceBook’s AI Research Director and a professor at NYU, described GANs
as “the most interesting idea in the last 10 years in [machine learning].” Yann LeCun, What Are Some
Recent and Potentially Upcoming Breakthroughs in Deep Learning? Quora (Jul. 28, 2016). GANs
are “adversarial” because two machines work one against the other, creating a constant feedback loop
that increases the quality of outputs. See AI Wiki, A Beginner's Guide to Generative Adversarial
Networks (GANs), https://skymind.ai/wiki/generative-adversarial-network-gan.
69
More specifically, GANs use an actor–critic model: as one machine, called the generator, gener-
ates new data instances, the other, the discriminator, “evaluates them for authenticity; i.e. the discrim-

The human cause 31
GANs can short-circuit the need for massive amounts of machine-learning, can produce much
better outputs and have “achieved remarkable results that had long been considered virtually
impossible for artificial systems.”70
More importantly, GANs are seen by some experts as “an
important stepping stone toward achieving artificial general intelligence [strong AI].”71
For
our purposes, to simplify, GANs are machines talking to machines and generating an output
for humans.
Machine-learning data can come from multiple sources, and AI machines often are “con-
tinually connected to the Internet and will continually take in new information and new pro-
gramming from multiple sources.”72
AI machines find correlations and detect new patterns in
data.73
Machines can for example correlate features such as voice to a series of characteristics
such as sexual and political orientation, certain diseases, and much more.74
Often, this predic-
tive ability of AI machines is “only” used with a commercial purpose, namely to determine
individuals’ preferences to sell them goods or services, but one can easily imagine far worse
scenarios.75
One risky feature of the use of AI machines in that context is that correlations are
usually based on data concerning the behavior of a given population but the impact is then
directed at individuals who may or may not actually fit the population’s behavioral patterns.76
4. APPLICATION TO PATENT LAW
AI machines “create a wide range of innovative, new, and non-obvious products and ser-
vices, such as medical devices, drug synthesizers, weapons, kitchen appliances, and [other]
machines.”77
There is little doubt that AI machines can help innovate and that they can produce
what looks facially like inventions as a matter of patent law.78
The question that this chapter
tackles is whether the law should provide patent protection for inventions in which human
involvement is not demonstrably and sufficiently present.
inator decides whether each instance of data it reviews belongs to the actual training dataset or not.”
Beginner’s Guide, ibid.
70
Jakub Langr and Vladimir Bok, GANs in Action: Deep learning with Generative
Adversarial Networks 3 (Manning Publications, 2019).
71
See ibid.
72
Jack M Balkin,, The Path of Robotics Law, 6 Cal L Rev Cir 45, 54 (2015).
73
Though not causal relationships. See Cary Coglianese David Lehr, Regulating by Robot:
Administrative Decision Making in the Machine-Learning Era, 105 Geo. L.J. 1147, 1157 (2017); and
Nick Wallace, EU’s Right to Explanation: A Harmful Restriction on Artificial Intelligence, TechZone
360 (Jan. 25, 2017), http://bit.do/Wallace_EU-Right-to-Explanation.
74
Ian Kerr Jessica Earle, Prediction, Preemption, Presumption: How Big Data Threatens Big
Picture Privacy, 66 Stan. L. Rev. Online, online: www.stanfordlawreview.org/online/privacy-and-big
-data-prediction-preemption-presumption/.
75
See ibid.
76
See Brent Daniel Mittelstadt et al., The Ethics of Algorithms: Mapping the Debate, Big Data
Soc’y, July-Dec. 2016, at 5. See also supra note 74.
77
Shlomit Yanisky Ravid Xiaoqiong (Jackie) Liu, When Artificial Intelligence Systems Produce
Inventions: An Alternative Model for Patent Law at the 3a Era, 39 Cardozo L. Rev. 2215, 2219–20
(2018).
78
See ibid.

Take DABUS, the test case in which the applicant named an AI machine as inventor.79
The European Patent Office, US Patent Trademark Office and, as of this writing, courts in
Australia, the UK and the US have found against the applicant and concluded that a human
inventor must be named in a patent application.80
Court decisions went the other way in South
Africa.81
Yet, stating that this question is a mere matter of naming a human overlooks the
actual normative issue. The underlying inquiry is whether patent law requires that a human be
the actual cause of an invention.
In functional terms, what are the legal requirements to be considered the “inventor”? Is
inventorship not necessary to claim that one should be named as an inventor? There are
several ways to address this line of inquiry, but the fundamental starting point is that this is not
a simple matter of applying and interpreting well-worn doctrines meant to separate ownership
claims to an invention to which multiple humans may have contributed. Under both UK and
US law as they now stand, there is little doubt that, under the current definition of inventor-
ship, a mere subjective belief that one is entitled to a patent as a basis to claim inventorship
is not the proper legal test.82
A more thorough rethink is in order because the question is not
the same as a multiple human inventor scenario. The novel question is: does a contribution
to the “conception” of the invention by a nonhuman entity legally qualify as inventorship as
a matter of patent law? Asking the question this way should not obscure the fact that the same
legal doctrines must also be tailored to novel types of human contribution to inventiveness,
such as programming and teaching AI machines, that were not part of inventive processes until
recently and which will gain prominence as AI machines get better at their job.
Under current law, the contribution of claimed inventors must be identified.83
This logically
presupposes that we know who, or what, the inventor is. As just noted, two major patent
offices and a court have found that one needs to identify one or more human inventors, while
other judges disagree.84
If one adopts the view that an inventor is a human notion under patent
law, then one or more humans, working in their unique way, must be causally related to the
invention.
A further, harder question will be to determine the role of patent incentives in that context.
Economic analyses will be useful but they won’t paint the full normative picture. Are we better
off as a society (here again using human progress, not disembodied technological change, as
a proper yardstick—how could it be otherwise?) issuing patents to machine-made inventions
79
On the USPTO, see Rebecca Tapscott, USPTO Shoots Down DABUS’ Bid for Inventorship,
IP Watchdog, May 4, 2020, online: www.ipwatchdog.com/2020/05/04/uspto-shoots-dabus-bid
-inventorship/id=121284/. For the EPO, see Bernt Hugenholtz, Daniel Gervais, João Pedro Quintais,
Trends and Developments in Artificial Intelligence: Challenges to the Intellectual Property Rights
Framework. Final Report (Nov 25, 2020), 100–04, online: www
.ivir
.nl/
publicaties/
download/
Trends
_and_Developments_in_Artificial_Intelligence.pdf.
80
Commissioner of Patents v. Thaler [2022] FCAFC 62; Thaler v Comptroller-General, UK Court
of Appeal, 21 September 2021; Thaler v. Vidal, 2022 WL 3130863 (Fed. Cir. 2022), respectively after
the reference to the UK decision. At the EPO, it seems that the naming requirement is a mere formality.
Patent offices rarely investigate actual inventorship.
81
See also chapters 20 and 23 in this volume.
82
See Ethicon, Inc. v U.S. Surgical Corp., 135 F.3d 1456, 1460 (Fed.Cir.1998) and Thaler, ibid.
83
See ibid.
84
See ibid. The argument that the technological neutrality required under TRIPS article 27.1 man-
dates patents on artificial inventions is unconvincing. AI technology remains patentable. It simply needs
to meet patentability criteria, just as computer software does.

The human cause 33
or not? Will doing so mostly accelerate innovation, or instead lead mostly to massive trolling?85
This is at bottom an empirical matter. Because AI has become a standard tool in many fields of
technology, empirical data about the production of new patents, the type of new technologies
produced, the employment of human researchers, and other relevant variables will become
available, which in turn should allow for a better framing of normative measures to tackle the
ongoing “cyborgization” of innovation. It may be that human scientists will spend less time
discovering and observing and more time interpreting data (and analyzing interpretations of
the data) provided by AI machines. The positives are easy to identify (new pharmaceuticals,
and so on); on the negative side of the ledger, however, how this might impact our quest to
understand nature and employment in applied sciences should be borne in mind.
As matters stand now, the chapter asserts, as a matter of policy, that those who claim we
(humans) would be better off by granting patents on nonhuman inventions have the burden
of proof primarily because this would result in applying a regulatory system meant to create
incentives for one type of activity (innovation springing from the human mind) to a different
type of activity (innovation from machines), which may not require the same set of incentives.
5. APPLICATION TO COPYRIGHT LAW
Should we protect artificial literary and artistic productions created without natural original-
ity, meaning productions the creation of which does not involve in a material way a human
creative process as cause?86
This would be a significant normative jump for, as Professor
Sam Ricketson—the co-author of the leading treatise on the Berne Convention—wrote, the
“need for authors to be ‘human’ is a longstanding assumption in national copyright laws.”87
Doctrinally, his observation seems entirely correct.88
Indeed, that assumption dates back to
well before the original (1886) text of the Berne Convention; it harkens back to the very roots
85
A troll in patent law is “a pejorative term describing a non-manufacturing patent owner who owns
one or more patents and asserts the patent(s) against alleged infringers, with a desire typically to obtain
settlement rather than actually trying any lawsuit.” Donald W. Rupert, Trolling for Dollars: A New
Threat to Patent Owners, 21 Intell. Prop. Tech. L.J. 1, 3 (2009).
86
This section provides a succinct overview of a more detailed argument presented elsewhere.
See Daniel Gervais, The Machine as Author, 105 Iowa L. Rev. 2053 (2020). See also Shyamkrishna
Balganesh, Causing Copyright, 117 Colum. L. Rev. 1 (2017) (developing a theory of authorial causation
connecting human agency to the expression embodied in a copyrighted work).
87
Sam Ricketson, People or Machines: The Berne Convention and the Changing Concept of
Authorship, 16 Colum. J. L. Arts 1, 8 (1991–2). Berne Convention for the Protection of Literary
and Artistic Works, Sept. 9, 1886, as revised at Paris, July 24, 1971, 828 UNTS 221 [hereinafter Berne
Convention]. The Berne Convention had 179 member States as of December 2020. The United States
became a party to the Convention on March 1, 1989. See World Intellectual Property Organization,
Contracting Parties: Berne Convention, www.wipo.int/treaties/en/ShowResults.jsp?lang=entreaty
_id
=
15 (accessed Dec. 15, 2020). The treatise referred to is Sam Ricketson and Jane C Ginsburg,
International Copyright and Neighbouring Rights: The Berne Convention and Beyond (2d ed,
2006).
88
An analysis of multiple national laws led another scholar to a similar conclusion. See Andres
Guadamuz, Artificial Intelligence and Copyright, WIPO Magazine (Oct. 2017) (“Most jurisdictions,
including Spain and Germany, state that only works created by a human can be protected by copyright”),
www.wipo.int/wipo_magazine/en/2017/05/article_0003.html.

of authors’ rights, as the word author comes from the Latin auctor, or originator.89
One could
go further. The entire path of copyright history follows the milestones of human creativity.90
Whether seen as a natural right—or even as a human right—or as an economic incentive,
historically the focus of copyright has unquestionably been on productions of the human mind.
If copyright had been designed as an investment protection scheme, or merely a scheme to
disseminate “things of value,” then the investment of publishers would have been sufficient.91
Be that as it may, we clearly are now faced with a new entrant in the battle for recognition
of authorship status. Does this new, intelligent “species” bend, or break, the normative arc of
copyright history? The first common law copyright statute—the Statute of Anne—provides
a good argument against protecting artificial productions.92
A set of arguments at the time
was that, if authors had an obligation not to write libelous or otherwise unacceptable content,
then authors should have a right in their writings.93
This created a normative link that seems
entirely convincing: if one is responsible for one’s writing, then one can legitimately ask for
a right in protecting moral or material interests in that writing.94
The argument rests on the
complementarity of responsibility and right, punishment and reward.95
A similar point can be
found in more modern work such as Foucault’s discussion of the persona of the author. He put
in parallel authorship and what he called “penal appropriation,” noting that “[t]exts, books,
and discourses really began to have authors […] to the extent that authors became subject to
punishment, that is, to the extent that discourses could be transgressive.”96
There is little doubt
in this author’s mind that owners and programmers of AI machines will distance themselves
faster than the speed of light if and when a machine they own or programmed produces infring-
ing or libelous content, though many of them of course will not hesitate to claim exclusive
rights if what the machine had produced is both non-infringing and commercially valuable.
89
For a detailed account of this evolution, see Gervais, note 86 supra.
90
See ibid.
91
See Mark Rose, Authors and Owners: The Invention of Copyright, 34–5 (Harvard Univ.
Press, 1993).
92
The Statute of Anne was the first common law copyright statute. Though adopted in England,
it served as a basis for the first state statutes and the first US federal copyright act in 1790. See Oren
Bracha, The Adventures of the Statute of Anne in the Land of Unlimited Possibilities: The Life of A Legal
Transplant, 25 Berkeley Tech. L.J. 1427, 1427–9 (2010).
93
See Rose, supra note 91, at 34–5.
94
Echoing the International Covenant on Economic, Social and Cultural Rights (ICESCR), art. 15(1)
(c), Dec. 16, 1966, 993 U.N.T.S. 3 (which recognizes “the right of everyone […] [t]o benefit from the
protection of the moral and material interests resulting from any scientific, literary or artistic production
of which he [or she] is the author”). See also Ben Saul, David Kinley and Jacqueline Mowbray,
The International Covenant on Economic, Social and Cultural Rights, 1226–9 (2014). As of
December 2020, the Covenant had 171 parties. The United States signed (but did not ratify) the Covenant
in 1977. See United Nations, International Covenant on Economic, Social and Cultural Rights, (Jan.
3, 1976), online: https://treaties.un.org/Pages/ViewDetails.aspx?src=INDmtdsg_no=IV-3chapter=
4

clang
=
_en (accessed Dec. 5, 2020). A human rights-based approach can inform parts of copyright
law, but in the past two decades copyright law at the international level has been shaped more by trade
agreements than human rights. See Daniel Gervais, Human Rights and the Philosophical Foundations of
Intellectual Property, in Research Handbook On Human Rights And Intellectual Property (C.
Geiger, ed), 89, 90–3 (2015).
95
See Rose, supra note 91, at 35–6.
96
Michel Foucault, What is an Author, in The Essential Foucault: Selections from The
Essential Works of Foucault 1954–1984, P. Rabinow N. Rose, eds (2003).

The human cause 35
This would amount to treating machines better than humans, letting them eat the proverbial
cake and have it too.
Why would artificial productions get a free pass? If the right/responsibility linkage served
as the justification for copyright for human authors, should it not be applied to machine
productions, and indeed to any other category of purported “author”? This would mean that,
once the machine production is not causally connected to one or more humans, there should
be no copyright in the production.97
There is an echo of this view in a resolution adopted by
the European Parliament on October 20, 2020, which recommended that regulators take into
account “the degree of human intervention [and] the autonomy of AI.”98
Ultimately, the risk of replacing humans in the act of creation, perhaps our noblest quest,
is the principal consideration. Protecting artificial productions which will come at no ongoing
cost and without the rights that a human creator would claim even after transferring her rights,
such as moral rights or rights reversion, means that market forces will lead to a fast replace-
ment of human creators whenever possible. It will also change what we read, watch, and listen
to. Machines will create based on existing material combined with data about what humans
are most likely to respond to, just like Facebook—or Meta—focuses on polarization instead
of informed discussion. The risk is that there will be more of the same, or worse. How that
can arrest human development is unclear, but the risk is nonetheless real. Letting machines
create the next waves of cultural development is fraught with existential risks. As with patents,
anyone can make this claim or the opposite, however, because it remains ultimately a matter
for empirical observation. Yet much of what can be lost in less than a generation may well
justify applying a version of the precautionary principle before we put copyright and the full
force of the market behind the replacement of human authors by machines.
6. USING CAUSATION TO SEPARATE HUMAN AND
MACHINE
Given the increasingly frequent conflation of human and machine contributions in the pro-
duction of creative and innovative content, it seems safe to predict that there will be more
productions that are both human and machine-made. As with works and inventions with
multiple contributors, courts may be asked to determine who, if anyone, the author and/or
right holder(s) should be. This is not the same inquiry as the one above, which was to decide
whether machine productions should be protected to begin with. This second inquiry is about
separating contributions, as might happen in any case of joint authorship or inventorship.
Courts will need analytical tools to separate human and machine contributions even if, against
the position taken in this chapter at least for copyrightable productions, the latter are ultimately
found to be protectable.
97
It is unnecessary, therefore, to delve more deeply into which human proxy should be, by legal
fiction, “selected” as the most appropriate right holder. To use the term in Ginsburg and Budiarjo, supra
note 50 at 439–42, the production is “authorless.” There is a specific concern in US law with respect to
the production of derivative works. See Daniel Gervais, AI Derivatives, 52:4 Seton Hall L. Rev 1111
(2022).
98
Intellectual Property Rights for the Development of Artificial Intelligence Technologies: European
Parliament Resolution of October 20, 2020 (2020/2015(INI)), online: www
.europarl
.europa
.eu/
doceo/
document/TA-9-2020-0277_EN.html.

Random documents with unrelated
content Scribd suggests to you:

The Project Gutenberg eBook of Arkielämää

This ebook is for the use of anyone anywhere in the United States
and most other parts of the world at no cost and with almost no
restrictions whatsoever. You may copy it, give it away or re-use it
under the terms of the Project Gutenberg License included with this
ebook or online at www.gutenberg.org. If you are not located in the
United States, you will have to check the laws of the country where
you are located before using this eBook.
Title: Arkielämää
Author: Elisabeth Maria Beskow
Translator: Anna Kurimo
Release date: March 3, 2019 [eBook #58999]
Language: Finnish
Credits: Produced by Juhani Kärkkäinen and Tapio Riikonen
*** START OF THE PROJECT GUTENBERG EBOOK ARKIELÄMÄÄ ***

Produced by Juhani Kärkkäinen and Tapio Riikonen
ARKIELÄMÄÄ
Kirj.
Runa [Elisabeth Beskow]
Suomentanut
Anna Kurimo

Sortavalassa, Karjalan kirjakauppa- ja kustannusliike, 1906.
1.
Katsoppas kuinka kirkasta vesi on! Voi aivan selvään eroittaa kivet
pohjassa! huudahti Erik ihastuksissaan ja kumartui veneen laidan
yli.
Hän ja hänen kaksoissisarensa Anna olivat juuri mieluisalla
soutoretkellään kahdenkesken.
Heidän isänsä oli Vesterlångan kirkkoherra. Pappila sijaitsi kauniilla
paikalla, Helsinglandin kauneimman järven, laajan Långsjön rannalla.
Lapset veneineen olivat kulkeneet etäälle kodista. He soutelivat
nyt hitaasti pitkin Braxenlahden tyyntä vedenpintaa. Ei näkynyt
jälkeäkään ihmisasunnoista. Lahtea varjostivat lehtevät rannat ja
salmi, joka johti siihen varsinaisesta Långsjö-järvestä, oli niin kapea
ja kiemurteleva, että sitä oli vaikea huomata. Sisarukset olivat
ikäänkuin eroitetut maailmasta. Ainoastaan yhdessä kohden, mistä
joku vuosi sitten oli metsä palanut, oli ranta aukea. Tuli oli kuitenkin
huomattu heti alussa ja rajoitettu ulkonevaan niemeen. Siellä
hohtivat nyt puolukat punaisina mustien kantojen välissä.
Oi Erik, ajatteles, että on jälellä vaan muutamia viikkoja, kunnes
sinun täytyy lähteä kouluun, sanoi Anna huoaten. Kuka sitten
soutelee ja leikkii minun kanssani? Uh! Älä nyt ajattele mitään
ikävää, vastasi poika. Keksikäämme sen sijaan jotakin oikein

hauskaa. Minä tunnen haluavani jotakin erityistä ja tavatonta juuri
nyt. Etkö sinä voisi ehdottaa jotakin?
Poimisimmeko puolukoita palaneelta niemeltä, ehdotti hän.
Ei, se on niin tavallista. Keksi jotakin parempaa!
Jospa menisimme maihin pienelle niityllemme ja rupeisimme
telmimään, ehdotti Anna edelleen.
Erik katsoi miettien mainittua niittyä kohti, joka pilkisti esiin
viheriänä ja houkuttelevana, tummien kuusten keskeltä.
No, olkoon menneeksi, sanoi hän iloisesti. Tunnen itseni
todellakin halukkaaksi telmimään.
Hän ojensi reippaasti käsivarsiaan, joiden lihakset pullistuivat
poikamaisen raittiina ja voimakkaina. Anna, joka istui soutamassa,
suuntasi veneen rantaa kohti. Erik seisoi keulassa, ketjut kädessä,
valmiina hyppäämään maihin, heti kun uskaltaisi.
Sitten kun lapset olivat sitoneet veneen rantaan, menivät he
pensaikon läpi ylös niitylle ja rupesivat telmimään. Se kävi siten, että
he ajoivat toisiaan takaa, reippaasti juosten ympäri, painivat ja
heittivät toisiaan kumoon. Tätä kutsuttiin kaksinkertaiseksi
telmimiseksi. Yksinkertaistakin telmimistä harjoitettiin väliin
vaihtelun vuoksi. Silloin juoksi kumpikin yksinään, heitti kuperkeikkaa
ja kieri kuin mieletön ikään lyhyessä ruohossa, potkien ja viuhtoen
käsillä ja jaloilla, mitä iloisimman innostuksen valtaama. Vähän aikaa
teuhattuaan väsyivät lapset leikkiin ja istuutuivat, läähättäen ja
nauraen vastapäätä toisiaan, maahan.

Kuinka hauskaa sentään on telmiä! Ei se ollut niinkään hullu
keksintö, sanoi Erik ja otti lakin päästään antaakseen raittiin
kesätuulen vilvoittaa otsaansa.
Niin, nyt on sinun vuorosi ehdottaa jotain hauskaa, vastasi Anna.
Erik mietti ja katseli ympärilleen. Silloin sattui hänen katseensa
suureen, litteään kiveen, joka oli niityn syrjässä, ja hänen silmänsä
alkoivat äkkiä säteillä.
Entäs jos leikkisimme pakanoita, sanoi hän. Kuinka me silloin
teemme? kysyi sisar innostuneena.
Me loihdimme. Niin pakanatkin tekevät ja käyttävät kiviä
alttareina.
Tuo kivi on mainio. Ja Erik juoksi ylös.
Mitä se loihtiminen on? kysyi Anna ja seurasi häntä.
Etkö sinä sitä tiedä? sanoi hän ylenkatseellisesti. Se on melkein
samaa kuin uhraaminen.
Mutta mitä me sitten uhraamme?
Täytyy keksiä jotakin, vastasi Erik ja istui aijotulle alttarilleen
miettimään.
Mutta sano, kenelle me uhraamme? kysyi Anna, joka seisoi
hänen edessään, kädet selän takana, täynnä innostusta.
Metsän jumalalle.
Onko semmoinen olemassa?

Ei, ei todellisuudessa. Mutta me leikimme, että semmoinen on,
ymmärrätkö?
Mutta kuinka voimme me uhrata hänelle, kun häntä ei ole?
Oh, sinä olet niin nenäkäs. Keksi sen sijaan jotakin sopivaa
uhrattavaa!
Soutaisimmeko palaneelle niemelle ja poimisimme puolukoita
metsän jumalalle, ehdotti Anna.
Sanottu ja tehty. He soutivat sinne ja poimivat puolukoita Annan
esiliinaan.
Mutta eikö ole väärin uhrata metsän jumalalle? Eikö se ole
epäjumala? Miksi emme yhtä hyvin voi uhrata omalle
Jumalallemme? kysyi Anna, kun he olivat saaneet tarpeeksi marjoja
ja soutivat takaisin niitylle.
Ei, vastasi Erik. Hän ei pidä siitä, että Hänen kanssaan
leikitään.
He astuivat maihin.
Poltammeko me puolukat?
Ei, meillä ei ole tulitikkuja, mutta tehkäämme vesiuhri. Nyt
olemme juhlallisia, sanoi Erik. Asetu tuohon ja pitele puolukoita
esiliinassasi, niin panen minä ne alttarille.
Anna totteli. Hänen sydämensä sykki juhlallisesta tunnelmasta,
johon sekaantui vähän pelkoakin, kun veli kaamealla äänellä lausui
salaperäisiä loitsuja hitaasti kourallaan ajaen puolukoita alttarille.

Hokus pokus filolokus, memsalis, pakaralis, huks taksura, labuna
kriks luki hän.
Oi, Erik, katsos minun esiliinaani! Mitähän äiti sanoo? sanoi
Anna, keskeyttäen juhlallisen loitsun. Mitä nyt! kysyi veli
kärsimättömästi, keskeyttäen lukunsa.
Katso, kuinka se on tahraantunut puolukoista, valitti tyttö.
Ja sinä itket sitten noin vähästä! Ole sen sijaan iloinen.
Näkyyhän, että metsän jumala on ottanut vastaan lahjamme, koska
tahrat ovat jääneet jälelle, lohdutti Erik.
Olisin kernaammin suonut, ettei hän olisi huolinut siitä. Nyt toruu
äiti minua, mutisi Anna.
Mutta Erik ei kuunnellut häntä, vaan jatkoi lorujaan, kunnes kaikki
puolukat olivat loppuneet esiliinasta.
Nyt ajamme me vettä niiden päälle, sanoi hän, ja sitten on uhri
valmis. Liitä kätesi yhteen maljaksi, niin käy se kyllä päinsä. Me
kaadamme seitsemän maljaa kumpikin.
He tekivät niin. Puolukat makasivat siinä niin kosteina, punaisina ja
tuoreina. Anna ei voinut olla ottamatta hyppysellistä niistä ja
pistämättä suuhunsa.
Anna, mitä sinä teet? huudahti Erik niin kiivaasti, että Anna
paralta oli marjat mennä väärään kurkkuun. Kuinka voit sinä ottaa
uhrimarjoista? Nyt en ihmettelisi vaikka metsän jumala tulisi ja veisi
sinut tuossa paikassa.

Anna raukka kalpeni säikähdyksestä ja katsoi tuskallisesti metsään
päin.
Minä otin niin vähän, änkytti hän.
Ei se auta, sanoi Erik ankarasti, nyt täytyy sinun joka
tapauksessa tehdä parannus. Mene hakemaan kuusen kuorta ja
aseta sitä marjojen päälle! Se pistelee, niin että se kelpaa
katumusharjoitukseksi.
Erik istuutui maahan, tyytyväisenä erinomaiseen päähänpistoonsa,
katsellakseen kuinka sisar täytti hänen käskyään.
Hyvä on, sanoi hän vihdoin, kun väsyi odotukseen ja nälkä
rupesi muistuttamaan, että päivällisaika oli lähellä, nyt luulen, että
metsän jumala on lepytetty ja sinä olet sovittanut rikoksesi.
He menivät alas rantaan soutaakseen kotiin. Anna koetti turhaan
saada pieniä pihkaisia käsiään puhtaiksi, Erik souti ja hän istui
perässä.
Kuules, sanoi tyttö lyhyen vaitiolon jälkeen, minä en pidä
metsän jumalasta, enkä tahdo enää uhrata hänelle.
Sinä teet tuhmasti, sanoi Erik lyhyesti. Miksi et tahdo uhrata?
Siksi, että se on epäjumala. Ei se varmaankaan ollut oikein tehty.
Mutta mehän leikimme vaan.
Niin, se on tietty, myönsi Anna, mutta ei ollut sentään oikein
tyytyväinen. Minä olisin tahtonut uhrata oikealle Jumalalle.

Sitten eivät puolukat olisi riittäneet. Se mitä uhrataan Jumalalle,
täytyy uhrata todellisesti, eikä vaan leikillä, ja sen täytyy olla
jotakin.
Mitä sitten on uhrattava Jumalalle, josta hän huolisi, ja miten se
on tehtävä? kysyi Anna. Erik mietti.
Niin, minä luulen esimerkiksi, että jos annamme rahamme
lähetykselle, niin uhraamme jotakin Herralle, jonka hän ottaa
vastaan, sanoi hän.
Minulla on kolme kruunua ja kaksikymmentä äyriä
säästölaatikossani, sanoi Anna vitkalleen, luuletko, että Hän
tahtoisi, että antaisin ne?
Kyllä luulen, että hän tulisi iloiseksi, jos tekisit sen.
Anna istui ääneti ja seurasi silmillään aironlapaa, joka nousi ja
laski vedessä. Se näytti niin kauniilta vaipuessaan lahden tyyneen
pintaan, jota myöten vene kiiti eteenpäin. Pian puikahti se lahdesta
varsinaiseen järveen ja Erikin voimakkaasti soutaessa suuntasi se
kulkunsa suoraan sitä rantaa kohti, missä Vesterlångan pappila
sijaitsi ja kylpi elokuun auringon valossa.
Ota kiinni nyt! Kas noin!
Erik veti sisarensa avulla kevyen veneen kauas maalle. Sitten
juoksi hän suorinta tietä keittiöön katsomaan mitä päivälliseksi
laitettiin.
Anna meni lastenkamariin ja otti laatikostaan pienen rasian, jossa
hän säilytti rahojaan. Hän laski kuinka paljon niitä oli.

Kuinka paljo minä annan? kysyi hän itseltään. Eiköhän
kaksikymmentä äyriä riittäisi, niin saisin pitää loput itse?
Mutta hän ei ollut tyytyväinen siihen. Antaisiko hän vaan pienen
osan Jumalalle ja pitäisi suuremman osan itse? Ei. Puoletko sitten?
Mutta eikö hän rakastanut Häntä tarpeeksi antaakseen Hänelle
kaikki? Jos hän pitäisi edes nuo kaksikymmentä äyriä itse ja ostaisi
niillä karamellia? Mutta ei, eikö hän voisi uhrata niitäkin Jumalalle?
Anna oli aikonut hankkia itselleen niin paljo näillä rahoilla. Mistä
olikaan hän saanut päähänsä tuon uhraamisen ajatuksen, joka ei
ollenkaan antanut hänelle rauhaa?
Mutta uskoisikohan Jesus, että hän rakasti Häntä, ellei hän tahtoisi
antaa Hänelle mitään? Reippaasti ja päättävästi tyhjensi Anna rasian
sisällön käteensä, jonka puristi lujasti kiinni. Sitten juoksi hän isän
työhuoneeseen, missä tiesi lähetyssäästölaatikon pienine, päätään
nyykäyttävine neekeripoikineen olevan hyllyllä. Hän nousi seisomaan
tuolille ja pudotti lantin toisensa jälkeen laatikkoon. Joka kerran kun
neekeripoika nyykäytti kiitoksensa, tuli Annan sydän kevyemmäksi.
Vihdoin hyppäsin hän tyhjin käsin lattialle ja tunsi itsensä iloiseksi
työnsä johdosta.
Kukaan ei ollut nähnyt hänen tekoaan, eikä hän kehdannut puhua
siitä.
Ainoastaan Erikin korvaan kuiskasi hän salaisuutensa.
Kuuleppas, minä olen uhrannut kaikki rahani Jumalalle, sanoi
hän.
Oh, teitkö todellakin sen? sanoi Erik pitkään ja katsoi
hämmästyneenä häneen.

Tein, enkä minä ollenkaan kadu, vakuutti hän.
Kas vaan, sinä, joka sentään tavallisesti olet niin ahne, sanoi hän
veljellisen suoralla tavallaan.
Olen niin kyllästynyt teidän puheisiinne minun ahneudestani, että
tahdon näyttää teille, etten olekaan niin ahne. Minä aijon tulla
anteliaaksi.
Sepä hauskaa. Tahdotko sitten alottaa antamalla minulle pienen
ruskean kivipallosi, ehdotti Erik.
Kyllä, sinä saat sekä sen, että sinisen, vastasi Anna jalomielisesti,
mutta kun veli todellakin otti hänen molemmat pallonsa, joutui hän
hämilleen.
2.
Omenat alkoivat kypsyä puutarhassa. Lapset eivät saaneet ottaa
puista, mutta pudonneet hedelmät kuuluivat heille. Sen vuoksi vallitsi
aamusin jalo kilvoitus ylösnousemisessa, sillä se, joka ensimmäiseksi
ehti puutarhaan, sai tavallisesti hyvän saaliin hedelmistä, jotka olivat
pudonneet yöllä.
Eräänä aamuna heräsi Anna jo kello viisi ja hyppäsi ihastuksissaan
vuoteelta. Nyt ei ainakaan ole kukaan ehtinyt ennen häntä! Hän
pukeutui kuumeentapaisella innolla ja riensi ulos.
Kaste kimalteli käytävillä, marjapensaissa ja puiden oksilla. Luonto
heräsi uuteen eloon auringon säteitten voimasta.

Ei ketään ollut näkyvissä. Anna kiiruhti heti astrakaanipuun
juurelle, mutta, ihmeellistä, siellä ei ollut ainoatakaan omenaa. Oliko
todellakin ollut niin tyyni yö? Hän astui kaalimaahan, omenapuun
alle, ja etsi olisiko mahdollisesti sinne pudonnut joku. Mutta ei
sielläkään ollut mitään.
Silloin kuului iloinen hyvää huomenta! Anna kääntyi ja näki Erikin
istuvan pienillä portailla aivan lähellä. Lakkia piteli hän polvillaan. Se
oli täynnä mitä kauneimpia omenoita. Takin ja housuntaskut olivat
arveluttavan pulleat. Oi kuinka hänen hymyilynsä Annan mielestä oli
kiusoittava! Hän kääntyi vihoissaan pois vastaamatta veljensä
tervehdykseen.
Oletko pahoillasi nyt kun petyit? kysyi Erik.
Anna polki jalkaa.
Aina sinä ehdit ensimmäiseksi. Vaikka minä nousisin ylös keskellä
yötä, olisit sinä sentään ehtinyt ottaa kaikki omenat, huudahti hän
itkuun pillahtamaisillaan.
Mutta Erik vaan hymyili.
Tule ja istu tähän portaille ja odota, niin ehkä vielä putoo joku
omena, sanoi hän, muista, ettei yö vielä ole loppunut. Makaahan
koko talon väki vielä.
Sitten istun minä ainakin sinun edessäsi, niin että minä ehdin
ensiksi, jos putoo, sanoi Anna yhä vieläkin vihaisena ja istui pari
porrasta alemmaksi veljeään.
Siten istuivat he hetkisen vaieten ja odottaen.

Sitten otti Erik hiljaa suurimman ja kauneimman omenansa ja
heitti sen sisarensa pään yli puuhun, niin että se putosi sen alle.
Nuolen nopeudella juoksi Anna ja otti sen.
Katso, Erik, minä sain sentään! Näin suurta ei sinulla olekaan,
huudahti hän riemuiten.
Ei sinun olisi tarvinnut olla niin vihainen, vastasi Erik vaan.
Anna istui taas paikalleen. Hän oli nyt leppynyt ja toivehikas. Pian
putosi omena taas alas, kuten äskenkin ja sitten vielä yksi. Silloin
rupesi Anna aavistamaan pahaa.
Eihän nyt tuule ollenkaan, mistä siis johtuu, että niin monta
omenaa putoo? ihmetteli hän.
Yöllä on tuullut ja ne ovat silloin alkaneet irtaantua, selitti Erik.
Mutta sinun lakkisi oli äsken niin kukkuroillaan ja nyt on se vaan
reunojaan myöten, huomautti hän epäillen.
Minun oli sääli sinua, kun et saanut yhtään, vaikka nousit niin
varhain ylös, tunnusti Erik.
Mutta sittenhän ne ovat sinun omeniasi?
Ei, ne ovat sinun nyt. Sinä saat ne.
Kiitos! Voi kuinka sinä olet kovin hyvä, sanoi Anna ja häpesi
sydämessään, että oli ollut niin vihainen äsken.
* * * * *

Oli joululoma luistinjäineen ja kelkkamäkineen. Veljet olivat tulleet
kotiin kaupungin koulusta ja riemu oli ylimmillään.
Eräänä aamuna, kun lapset heräsivät, oli koko Långsjön pinta
yhtenä ainoana kirkkaana iljanteena. Heti aamiaisen jälkeen menivät
he nauttimaan siitä. Pojat, Kurt ja Erik, olivat taitavia luistelijoita ja
Annakin oli jo hyvällä alulla. He olivat luistelleet hetkisen, kun äiti tuli
alas rantaan heidän pienen, nelivuotisen sisarensa kanssa.
Katsokaas Evaa, huudahti Kurt ja riensi pienokaista vastaan.
Eva oli kaikkien lemmikki. Kurt kohteli häntä kuin pientä
kuningatartaan ja tämä ottikin vastaan hänen suosionosoituksensa
pienen kuningattaren tyynellä arvokkaisuudella.
Hankittiin kelkka ja Eva istui siihen. Veljet vetivät kilvassa kelkkaa,
äidin kävellessä vieressä.
Pikku Eva taputti ihastuneena käsiään. Punaiset posket kävivät
talvikylmässä vieläkin punaisemmiksi. Sitten sai hän juosta jäällä
veljien pidellessä kiinni molemmista käsistä. Jos hän kaatui,
kannattivat veljien voimakkaat käsivarret häntä, niin ettei hän
koskaan loukannut itseään. Hän nauroi ja huusi ihastuksissaan.
Hetkisen kuluttua tuli isä käyden. Silloin riisti Eva itsensä irti ja
koetti juosta häntä kohti, mutta luiskahti ja olisi kaatunut, ellei isä
samassa olisi ottanut häntä syliinsä. Hän nosti hänet korkealle ylös ja
suuteli kerran toisensa perään hänen pyöreitä poskiaan. Jos joku
lapsista oli hänen erityinen suosikkinsa, niin oli se pieni,
hellämielinen Eva. Kun oli aika tämän mennä sisään, piti isä häntä
sylissään ja kantoi hänet jään yli kotiin.

Mikä sinun on, Anna? Miksi olet niin nyrpeällä nenin?
Se oli Erik, joka kysyi. Hän oli kaivannut sisartaan ja etsinyt,
kunnes oli löytänyt hänet yksinäiseltä lahdelta, jossa hän istui
rannan kivellä ja hakkasi luistimellaan jäätä.
En minä ole nyrpeissäni, vastasi hän.
Kyllä sinä olet. Mikä sinua vaivaa, sano?
Mutta Anna ei voinut sanoa sitä, hän ei tietänyt sitä itsekään. Hän
ei ymmärtänyt, mikä se tunne oli, joka oli vallannut hänet,
nähdessään, kuinka veljet olivat hyväilleet Evaa. Tämä tunne oli
kohonnut huippuunsa, kun isä oli tullut ja ottanut pienokaisen
syliinsä ja suudellut siten kuin Anna ei koskaan muistanut hänen
suudelleen häntä. Kaikki pitivät Evasta, ei kukaan hänestä. Mistä se
johtuu? Joku aika sitten oli pappilassa ollut vieraita. Silloin oli Anna
kuullut erään tädin sanovan toiselle: Tyttö parka, hän on kovin
ruma. Ja sitten olivat he katsoneet häneen. He olivat puhuneet
hiljaa, mutia Annalla oli tavattoman tarkka kuulo. Nämä sanat olivat
syöpyneet hänen sydämmeensä, Eva oli sillä kertaa ollut sisällä
yht'aikaa ja häntä oli ihailtu ja hyväilty.
Olenko minä kovin ruma, Erik? kysyi hän nyt istuessaan kivellä.
Ruma, toisti veli ja katseli tutkivasti sisartaan, en minä tiedä, en
ole koskaan ajatellut sitä. Niin, sinä olet kalpea, sinulla on viheriät
silmät ja karkeat hiukset ja suu, jossa ei parhaillaan ole juuri
ollenkaan hampaita. Kyllähän sinä olet ruma, mutta mitä se tekee?
Älä huoli siitä. Oletko nähnyt kuinka kauniita kahdeksannumeroita
minä osaan luistella? sanoi hän ja näytti uutta taitoaan.

Anna katseli häntä hajamielisenä.
Minä en osaa vielä luistella takaperin, sanoi hän
välinpitämättömästi.
Hän oli Erikin mielestäkin ruma. Hänen rohkeutensa aleni vieläkin
yhden asteen.
Sitten opetan minä sinua, sanoi Erik vilkkaasti. Me pitelemme
toisiamme käsistä, sinä luistelet taaksepäin, minä eteenpäin, niin opit
sinä pian.
Sanottu ja tehty. Anna innostui niin tähän uuteen urheilun
haaraan, että hän aivan unohti huolensa ja kateuden haltiatar pakeni
tällä kertaa hänen lapsensydämestään.
3.
Neljännestunnin matkan päässä pappilasta oli Lippuvuori, joka sai
nimensä pienen lipun muotoisesta tuuliviiristä, joka oli pitkän tangon
päässä vuoren huipulla. Ylt'ympärillä oli kaiken suuruisia kiviä ja
niiden kanssa leikkivät lapset rakentaen niistä tupia, muureja,
holvikäytäviä ja pihoja. Tuulen humina kuusissa ja lehtipuitten
kahina sekaantui pienen rautalipun narinaan ja säesti heidän
puhettaan.
Kurt teki päätöksen. Hän tahtoi rakentaa majan, niin suuren, että
sen sisällä mahtui olemaan. Toiset lapset hämmästyivät ihastuksesta
ja ihmettelivät miten se onnistuisi. Kurt valitsi sopivan paikan
keskellä Lippuvuorta ja alotti työnsä parin pojan avustamana. Erikkiä

ja Annaa ei hyväksytty työntekijöiksi. He olivat naisia ja lapsia, selitti
neljäntoista vuotias Kurt ylenkatseellisesti, toisten suureksi
mielipahaksi. He leikkivät etempänä männynkäpykarjoineen ja
tekivät yhä edelleen pieniä käytäviä ja navetoita, parhaiksi sopivia
kävyille, samalla kun he kateellisesti katselivat, kuinka Kurtin
kivirakennuksen seinät kohosivat.
Kuuleppas, sanoi Erik.
Mitä? kysyi Anna.
Jospa mekin rakentaisimme itsellemme majan; jonka sisällä voi
olla?
Annan silmät säteilivät.
Oi, tehdään se! huudahti hän.
He menivät heti valitsemaan paikkaa ja löysivät pian tarkoitukseen
sopivan muutaman samasta juuresta kasvaneen kaksoismännyn
vierestä.
Katso, sanoi Erik, osoittaen näitä, nuo saavat olla mukana
seinässä.
Ne tukevat koko majaa.
Anna ihaili hänen viisauttaan, ja sitten rupesivat he rakentamaan.
Kurt ja hänen auttajansa huomasivat heidän aikeensa ja nauroivat
ivallisesti.
Tuosta ei koskaan tule mitään, sanoivat he ja rakensivat
itsetietoisina paremmuudestaan palatsiaan. Mutta Erik ja Anna olivat

kestäviä. He tekivät innokkaasti työtä ja vierittivät rakennukselleen
niin suuria kiviä, että he itsekin ihmettelivät voimiaan.
Aurinko paahtoi hikiset lasten kasvot ruskeiksi, pienet kädet tulivat
multaisiksi ja kosteiksi sammalpeittoisista kivistä. Sitten laskettiin
perustus ja pieni kivimuuri rupesi kohoamaan kaksoismäntyjen
juurelle.
Kun päivälliskello soi, kokoontui nälkäinen, väsynyt, mutta innokas
seurue pöydän ympärille. Erik ja Anna kertoivat säteilevin silmin
suuresta aikeestaan.
Ja minä jaksan vierittää näin suuria kiviä, huusi Anna, osoittaen
mittaa käsillään.
Mutta minä vielä suurempia, lisäsi Erik innokkaasti.
Kunhan te vaan ette liiaksi ponnistele voimianne, varoitti äiti
huolestuneena.
Oh, älä ole levoton, sanoi isä hymyillen, se tekee vaan heille
hyvää. Katsoppas ruokahalua!
Äiti katsoi rakkaittensa ruoka-annoksia ja hymyili hänkin.
Rakennusintoa jatkui muutaman päivän, mutta sitten rupesi Anna
väsymään. Hän huomasi, että Erik sanoi minun metsämökkini
meidän asemesta, ja rakennettaessa kävi yhä selvemmin ilmi, että
hän piti itseään päällysmiehenä ja Annaa käskyläisenä. Hän määräsi
ja sisar sai tehdä sen mukaan. He eivät rakentaneet yhdessä yhteistä
mökkiä, vaan Erik rakensi ja Anna oli vaan apulaisena. Se ei Annan
mielestä ollut enää ollenkaan hauskaa. Erikille ei hän sentään

Welcome to our website – the perfect destination for book lovers and
knowledge seekers. We believe that every book holds a new world,
offering opportunities for learning, discovery, and personal growth.
That’s why we are dedicated to bringing you a diverse collection of
books, ranging from classic literature and specialized publications to
self-development guides and children's books.
More than just a book-buying platform, we strive to be a bridge
connecting you with timeless cultural and intellectual values. With an
elegant, user-friendly interface and a smart search system, you can
quickly find the books that best suit your interests. Additionally,
our special promotions and home delivery services help you save time
and fully enjoy the joy of reading.
Join us on a journey of knowledge exploration, passion nurturing, and
personal growth every day!
ebookbell.com

Research Handbook On Intellectual Property And Artificial Intelligence Ryan Abbott

More Related Content

Similar to Research Handbook On Intellectual Property And Artificial Intelligence Ryan Abbott (20)

Recently uploaded (20)

Research Handbook On Intellectual Property And Artificial Intelligence Ryan Abbott