Artificial Intelligence + Robots

Weekly Barometer 25 janvier 2012
ARTIFICIAL INTELLIGENCE & ROBOTICS
OCTOBER 2015
ATONRÂ PARTNERS SA 12, Rue Pierre Fatio – 1204 GENEVA – SWITZERLAND - Tel: + 41 22 310 15 01 http://www.atonra.ch

THE 3RD AND FINAL STAGE OF TECHNOLOGICAL EVOLUTION
 Over the years, we have extensively covered the ongoing technological evolution in our
Weekly Barometers. But when meeting with investors, we often face the same questions
and notably: What is Artificial Intelligence really about?
 Artificial Intelligence is about computer systems doing what is usually performed by the
human intelligence
 That’s why we believe we have entered into in the 3rd and final stage of the technological
evolution in which:
 automation allows computer systems to control other machines
and, more importantly, to replicate the five human senses
 While AI is not a new concept (1930s), what really makes
it interesting today is the conjunction of different factors:
 rising computing power, reliable networking, more R&D
efforts from universities and labs, advancement in new
materials and, last but not least, the continued need from
corporations to improve productivity
 However this is a development that is just beginning and before we can have all of the five
senses truly replicated, many years (if not decades) of R&D are still ahead of us
 In addition we would be wary of pockets of hype: some innovations are called AI but have
nothing to do with it as they rely on existing technologies and mainly on “data extraction”
2

THE DIFFERENT STAGES OF TECHNOLOGICAL EVOLUTION
 The different stages of technological evolution can be summarized as follows:
 TOOL (prehistoric man) → This was the 1st stage of technological evolution
 MACHINE (the British industrial revolution) → Represents the 2nd major stage of tech
evolution
 AUTOMATION is the 3rd and final one, with the removal of the human control from the
machine
 Evolution (as per the Darwin theory) is nature’s modification through small variations over a
long period of time, but recent technologies modified:
 the speed at which changes have been taking place since the 2nd half of the 20th century
 the potential to change humans’ biological boundaries
 Nothing can stop the pace of innovation (it’s in human nature) and today’s and tomorrow’s
techniques are only giving a boost to the ongoing pace of innovation
 A sole item is not responsible for the take-off of AI and Robotics, rather the combination of
several techniques. But two variables are acting as very important catalysts:
 Internet → Plays a very important role as an information and collaborative tool
 Productivity → This is the real engine (together with competition) for corporations to stay
ahead (and keep alive) vs. the competition
3

WHERE ARE WE HEADING?
 In order to better understand and possibly answer this question we must first understand the
human nature itself
 According to Prof. Nayef Al-Rodhan, a world recognized neuroscientist who heads the
Geopolitics and Global Futures Program at the Geneva Centre for Security Policy
 Human Nature is emotional and amoral
 if humans have to deal with hostile environments, moral norms won’t prevail
 a minimum of well-being in addition to a common ownership, must be earned
 AI tries to master the methods of biology (including the human intelligence), one of the best
examples being the Blue Brain Project in Lausanne, Switzerland, which got $1.3B in funding
 the idea is that we could have (not before some decades) a merger of biology and human
technology that would expand human intelligence
 While in the Western world this is seen as worrisome
(the Lausanne example shows how mindsets evolve),
Japan has a different view
 Japanese believe that everything (and every object)
has a spirit, explaining why the country is one of the
leaders in the Robotics space
 and notably the first country to have come out with a
robot (developed by Mr. Hiroshi Ishiguro) that looks and
behaves as an human
4

WHERE ARE WE HEADING AND WHERE DO WE CURRENTLY STAND?
 According to Vernor Vinge “We are on the edge of change comparable to the rise of human
life on Earth”
5

THE MERGER OF TECHNOLOGY AND HUMAN INTELLIGENCE
 This picture is self-explanatory…From “The Singularity is Near” by Ray Kurzweil
6

DO WE REALLY NEED TO WORRY ABOUT AI AND ROBOTS?
 A serious debate has been taking place in the last two years about the possible impact of AI
and Robots on many fields of the economy and on the role of humanity
 Mr. Stephen Hawking warned the world about the possible dangers of AI stating that:
 “the development of full artificial intelligence could spell the end of the human race”
 Elon Musk (founder of Tesla) called AI as:
 “our greatest existential threat” and “summoning the demon”
 Mr. Bill Gates also stated that:
 “I don’t understand why some people are not concerned with AI”
 The irony here is that these people all use or stand to benefit
from AI:
 Mr. Hawking uses an AI system that learns how the professor
thinks and then suggests words he might want to use next
 Mr. Musk’s Tesla cars already come with a built-in driver assist
system and his company is almost ready to have cars driving without human assistance →
Mr. Musk was also one of the early investors in DeepMind (bought by Google)
 and Microsoft is currently developing the Personal Agent project, a system which helps
people with their memory, attention and focus
7

A “SOMEWHAT” DIFFERENT VIEW FROM GOOGLE AND DEEPMIND
 Since Google hired Ray Kurzweil in 2012, as director of engineering focused on machine
learning and language processing,
 The company has made many interesting acquisitions and notably Jetpac and DeepMind
 Jetpac is a system that looks for visual cues to determine contextual information about the
area where a photo was taken, and reviews that visual information to determine what’s
actually happening in a given location
 DeepMind (a UK start-up) on the other hand is much more intriguing and vague but tries to
build a system that thinks…
 … based on deep learning (or artificial neural networks), in which the function of the brain is
mimicked, and tasks can be solved without prior programming
 But what’s even more interesting is that Google agreed to establish
 an ethical board to ensure DeepMind’s artificial intelligence technology isn’t abused
 and it is believed (not confirmed by Google) that Google already uses a handful of DeepMind
features for its search engine
 One interesting feature of the DeepMind acquisition is that the founder (Mr. Demis
Hassabis) has developed much of his interest by developing computer games
8

VIDEO GAMES ARE A MAJOR FIELD OF EXPERIMENTATION FOR AI
 On its website, Microsoft Research says the following about AI: “Let us begin with a
provocative question: In which area of human life is AI currently applied the most? The
answer, by a large margin, is Computer Games.“
 Indeed, the video game industry can be seen as a pioneer in
the field of AI:
 most games have been using AI to generate intelligent
behaviors in non-player characters and opponents
 since the industry’s take-off in the 70s and 80s and its most
basic characters
 until today’s realistic characters
 Today, the relation between AI and video games is as strong
as ever:
 while Microsoft talks about “a mutually beneficial combination”.
On the one hand, AI technology can provide solutions to an increasing demand to add
realistic, intelligent behavior to the virtual creatures that populate a game world
 on the other hand, as game environments become more complex and realistic, they offer a
range of excellent testbeds for fundamental AI research.”
 Google, when experimenting AI, does it with video games. Notably, Google’s DeepMind AI
software played retro Atari games. Out of 49 games, Google's AI outperformed humans in 29
games and the best-known algorithmic methods in 43 cases
9

VIDEO GAMES AND AI: VR COMING NEXT
 Virtual reality, the next evolution in gaming, will only strengthen the interactions between AI
and video games
 While research into virtual environments and research into AI have largely been carried out
separately, by different groups of people with different objectives, convergence is now
apparent
 as game developers now have to produce intelligent digital environments
 Interestingly, Facebook lists both AI and VR among its major initiatives and is expected to
make a major push into VR next year when it releases the Oculus headset
 Zuckerberg: “we’re working on VR because I think it’s the next major computing and
communication platform after phones”
 In conclusion, video games are a major
area of experimentation for AI and what
happens in this industry should be watched
closely
10

WHERE’S THE “HOT MONEY” GOING?
 With some high-profile acquisitions made in the past two years by Google and others such
as Apple, Facebook, Baidu etc.
 And dramatic improvements in technologies and materials, and the constant need by
corporations to be more efficient and competitive in an ever increasing globalized world
 a lot of VC money has started to flow into this sector
 ranging from drones to medical and non-medical robots, sensors, computer vision, speech
recognition, energy savings…
 According to a Bloomberg study, machine learning and deep learning saw a 20-fold increase
in VC money from 2010 levels (admittedly from very low levels and mainly into startups)
 This is particularly important for the future of AI & Robotics, as competition among startups
and bigger companies is to drive future growth which:
 is to drive even more innovation (an innovation virtuous cycle is thus in sight) and more
products and services in the market that need to be more effective and valuable
 We think that interest is likely to get much stronger in the coming years, as all the economic
sectors are to be impacted including financial, retail, marketing, consumer electronics,
accounting…
 With this in mind we believe that we are only at the beginning of the cycle, which is likely to
last for the next few years
11

MONETIZING AI: TRANSFORMING DATA INTO REVENUE STREAMS
 Data is growing exponentially, supported by the adoption of cloud, mobile and social
networking and soon IoT, with IDC/EMC claiming that the world’s information is doubling
every 18 months
 As such Artificial Intelligence systems or applications associated with AI are set to become
more accurate and more mainstream in the next few years as:
 AI systems are fed with data and do learn from data (any kind of data including unstructured
data) → The more the data, the better DeepLerning techniques are to become
 Firms are incentivized to get access to new revenue streams and/or better efficiencies by
investing in data mining and data-analysis systems, however such systems:
 rely (at the design/programming level) on people with very high skills who are expensive, as
such only accessible to a limited number of companies
 and this largely explains why the bulk of data (90%) is still unused
 During the last few years, very successful SaaS business models (such as Salesforce in
CRM) took off, as they were able to offer products with a clear utility function and accessible
to the most at a competitive price
 We believe that the best business models around tomorrow’s applications based on AI (such
as predictive modelling, decision analysis etc.) are likely to be SaaS-based
 allowing corporation to boost their sales and reduce costs
12

MONETIZING AI: MAIN APPLICATIONS IN THE CORPORATE WORLD
 While analytics software have been historically used for risk assessment purposes,
 notably in the insurance/banking sector by helping to define, measure, and manage risk
(credit scoring…)
 The combination of Big Data and algorithms makes them much more powerful today and
enable a new range of opportunities
 Decision Analytics notably is rising fast:
 these tools help to price product and services, detect and prevent fraud, predict future
losses…
 In other word, they power the decision-making process in almost all industries and functions
 Specifically, marketing applications are numerous:
 decision analytics software tracks and analyzes marketing campaigns, KPIs and customer
behavior to identify untapped marketing opportunities
 and can make recommendations on corporate marketing programs and on how to acquire
and retain customers
13

DATA STANDS AT THE CORE OF AI SYSTEMS
 Big Data (and large data sets) has been a boon for AI, as the more the data the better AI systems
perform, as some of those systems use statistical and probabilistic models
 When we speak to people working in the AI (or cognitive technologies) space, one of the main
recurring problems is data availability
and accessibility
 Most of the techniques used in AI try to discover patterns
from large data sets
 for example in NLP (natural language processing)
large data sets are needed to have an automatic
system that extracts the main topics from the
document and self generate a correct and
readable text
 another example comes from Computer Vision,
where a large database of images is needed as a
way to train the model to recognize persons,
shapes, colours etc. and improve for example
the ability by such systems to classify objects
 If data is not available, most of the algorithms developed are useless, as they can not train
themselves in order to suggest better results
14

DO THOSE CONTROLLING DATA STAND TO CONTROL EVERYTHING?
 Lately we held discussions with one of the world’s prominent professors in the field of AI
 he told us that he had a great patented algorithm which could match people’s wishes in terms
of hotel accommodation with hotels’ own service offerings and availability
 When his lab asked for more data from the likes of TripAdvisor, he was told that these
companies were thinking to develop such systems by their own
 and that they would not give away nor sell their data
 While there’s a ton of data available nowadays, with this example we have shown that in the
future this data might not be accessible to all
 And those corporations with huge data sets (such as Google, Facebook, Amazon etc.)
coming from from billions of Internet-connected devices are the ones to stand out and the
likely winners
 unless new business models (I give you access to my system which has a utility function in
exchange for data collection) are invented by new startups
 Our key questions (we have no answers yet) are the following:
 why are people so silly to give away their data (which has a great value) for nothing?
 why the concept of a “free” internet is only applicable to a few?
 what if the Telecom Operators (those transmitting data and as such sitting on a golden mine)
are to take control of this data? We think this explains why Facebook and Google are to
develop their Telecom infrastructure even more
15

MONETIZING AI IN THE REAL WORLD: ROBOTS AND IoT
 While AI is already extremely powerful, it is essentially a geek stuff for now
 sophisticated software enabling to extract and analyze data and powering the decision-
making process
 If AI wants to show off its abilities at some point and become more mainstream, it will have
to interact with the real world
 Robots will obviously be a major application
 but AI can also interact with the real world through the RFID (and IoT) technologies, which we
have long been highlighting in past Weekly Barometers
 As no one in the market has really the financial power of Google, those willing to enter the AI
and Robotics race are obliged to think market-wise
 and to come up with products that generate cash-flows, which could eventually be reinvested
into R&D
 There’s a number of start-ups and listed companies currently working in AI for a myriad of
applications (from advertising to security, defense etc.) and this is a field that is likely to
continue to see a lot of inflows in the next few years
 due to the potential of the addressable market
 in addition to the discoveries it might bring and the problems it might solve
16

RISE OF THE (COLLABORATIVE) ROBOTS
 As commented in previous pages, robots and, specifically, collaborative robots
 are likely to be of AI’s main applications
 We have long been highlighting the promising outlook for the industry, stating that the next-
generation of robots has a compelling ROI proposition
 keeping in mind that robots can work 24/7, never call in sick, never go on strike…
 Current generations of robots are easier to program and come with features such as vision
and touch technologies
 making them a first choice for all the heavy lifting and dangerous tasks and almost all the
precise work
 Collaborative robots can work alongside humans without a safety cage, meaning they can
easily be deployed within a factory
 And last but not least,
 collaborative robots are much less expensive than
traditional robots, suggesting that deploying a fleet
of robots for a large industrial group is quite easy
 $30-40k on average vs. $50-250k for traditional
industrial robots + high integration costs
17

NEW PRODUCT CYCLE TO PROPEL THE ROBOTIC INDUSTRY
 Importantly, we believe that we are at the early stage of a new product cycle as most legacy
players (ABB, Fanuc, Kuka…) and newcomers (Universal Robots, ReThink Robotics)
are expected to ship new products this year or next
 while privately-held ReThink Robotics is now shipping Sawyer at less than $30k (not far from
Universal’s price point)
 ABB has announced YuMi, priced at $40k
 and Kuka, which currently retails a collaborative robot which we do not view as competitive
(price point above $50k), is expected to launch a “low-cost” version in 2016
 These new products and price points come at a time when rising labor costs in emerging
markets and more specifically in China
 will probably spark some factories to switch from humans to robots,
 something Foxconn has talked about for a pretty long time
 Hence, we believe these numerous product introductions and lower price points will
significantly expand the market and suggest that current expectations leave room for upside
18

LOW EXPECTATIONS LEAVE ROOM FOR MAJOR UPSIDE
 The International Federation of Robotics expects only mid-teens shipment growth for the
robot market over coming years (with logical price erosion)
 That sounds like a very conservative forecast given the +29% shipment growth recorded last
year and the new product cycle
 To put this another way, there are roughly 1 billion people performing manual labor around
the world:
 assuming robots replace only 5% of these jobs in 10 years, the robot market would reach
50M units or 5M units per year on average, compared to 229,000 in 2014…
19

“SOFTWARE” TO RESHUFFLE THE CARDS IN THE ROBOT INDUSTRY
 Contrary to what most people think
 we believe the barriers to entry are pretty low in the Robot industry
 A robot is basically a combination of
 steel, servo-motors, sensors, tools to perform tasks (such as cutting lasers and other
equipment) and software (OS, machine vision…)
 Considering that steel, servo-motors, sensors and tools are widely available
 new players can easily step in and develop their own robots, as illustrated by the emergence
of companies such as ReThink and Universal Robots in the last few years
 Even if we admit that gaining credibility and customers is not an easy task (as in most
businesses), in our view the advent of collaborative robots could dramatically reshuffle the
cards in the robot industry, as software becomes a critical component
 as an illustration, ReThink Robotics’ first robot, Baxter fell behind Universal's model on speed
and precision, according to Frank Tobe of industry site The Robot Report and as a
consequence, Baxter shipments were rather disappointing
 but Rethink made major software improvements on Baxter during 2014, with the latest
software release doubling the robot’s speed and making it better equipped for higher-
performance tasks
 It clearly suggests robots are not a simple hardware business anymore, but a mix of
hardware and software that users can upgrade over the air like an iPhone or a Tesla
20

SUPPLIERS APPEAR AS A SAFER BET THAN ROBOT MANUFACTURERS
 In view of the disruptions seen in many industries (mobile phone business, car industry) in
recent years due to the rising importance of software
 we wouldn’t be surprised to see robot start-ups such as ReThink Robotics gain significant
share at the expense of incumbents
 Notably as these companies have the backing of well-known venture capitalists and tech
entrepreneurs (ReThinks has notably Bezos Explorations among its shareholders) who will
connect them with the right software engineers and skills
 Hence, we would be wary of most incumbent players,
notably those that already enjoy high margins (such
as Fanuc, 30-40% margin)
 And would favor suppliers, as they will clearly
benefit from the industry shipment growth
whoever wins the robot race, in the following fields:
 providers of sensors, vision and range-finding
technologies such as Keyence
 providers of motors
 providers of cutting lasers and tools for robots
 vendors of testing and measurement equipment
21

ULTIMATELY, THE TECH GIANTS COULD PLAY A MAJOR ROLE
 In a near future, home and medical robots are likely to be all the rage…
 and to be the new battlefield of tech giants if one assumes that this
business will reach the unit shipments and/or margins traditionally
achieved by device makers and Medtech companies
 Confirming our view that software could make the difference in the
robot industry
 Google has bought a flurry of robotics start-ups in the last
24 months, among which Boston Dynamics, Redwood Robotics,
Meka Robotics or Japan-based Schaft
 and has filed a patent for developing “methods and systems for
robot personality development” that will allow to download from
the cloud personality traits for robots
 By combining its robotics assets with AI (in which it is expected to
play a major role), Google could have a cutting-edge in next-gen robots
 Other tech giants do not remain idle: While Softbank has developed the
Pepper robot and an OS that controls robots in the same way Google’s
software runs smartphones and tablets
 Sony has been working on a next generation mechatronic device that
is seen as a tool for future seniors and/or nursing home facilities
 Amazon has a clear interest in robots as it needs to automate as much as possible its
warehouses while Apple needs to find significant growth drivers
22

M&A IN SIGHT
 Against this backdrop, we believe that the small players are obvious targets either for large
incumbents such as ABB or tech giants
 Following the Universal acquisition by Teradyne, a US-listed provider of test equipment for
the semi and electronics industries
 ReThink could be next on the M&A list even if some of its backers (notably Goldman Sachs)
will probably prefer an IPO
 Kuka could be the other M&A target as its relatively small market cap (EUR2.5bn) that could
easily be swallowed by a large group
A clear preference for suppliers over robot makers
 Given the competitive risks we identify in the robot
manufacturing business, we have a preference for
suppliers over robot makers
 as the former will clearly benefit from any
industry shipment growth while the latter
could be affected by volume and pricing pressures
 And among robot makers, we tend to focus on the
smaller names such as Kuka
 that could be taken over and on medical robot
makers, which are less exposed to competitive
threats
23

IoT & RFID: EVERY SINGLE OBJECT TO THE INTERNET AND TO… AI
 As commented previously, we believe the RFID and IoT technologies will enable AI to
interact with the real world
 RFID enables data transmission between an object and a reader and is seen as one of the
enablers of the IoT revolution
 Historically, RFID labels have been used to make the tracking and location
of items much easier and more efficient than barcode technology,
specifically in the retail sector
 Interestingly, many new applications are emerging for RFID tags as the
technology improves
 Notably, RFID tags can extract and record data from a variety of
items such as:
 concrete, with Holcim/Lafarge including tags in its products to track them and,
more importantly, to collect data on the material qualities over time
 temperature-sensitive products (pharmaceuticals or biomedical items)
24

IoT & RFID: WHEN, TOMORROW, NO HUMAN DECISION IS NEEDED
 In a near future, it is highly likely that the data collected through RFID will be processed by
AI software that will take decisions based on its conclusions
 Just a few examples:
 based on real-time information of a
store network’s inventories, AI will
decide to increase or reduce factory
production/shipments
 in hospitals, the combination of AI and RFID
will enable care tracking and, specifically, more
effective and more timely care provision
 in your (smart) home, RFID will detect when
you are running low on your favourite products
(beverages, food, dishwasher detergent…), and AI will detect and order them without your
intervention in quantities that are consistent with your consumption habits
 It looks like the Amazon dash button but goes one step further as there is no need of human
intervention
 Hence, smart homes will not only be about a set of smart devices, but about an AI ecosystem
that adapts to changes in behavioral patterns, anticipates needs and adjusts home
automation accordingly
25

ARTIFICIAL INTELLIGENCE & ROBOTICS 26
THE HISTORY OF AI

OUR INVESTMENT PRODUCTS AND PROCESS
 On July 24 2015, we launched an AMC (actively managed certificate) on Innovative
Technologies
 the certificate is focused on innovations that we believe will reshape the global economy in
the next few years, such as robots, artificial intelligence, the internet of things, cyber security,
graphene and many others…
 Earlier this month, we launched another AMC totally focused on AI and Robotics
 We clearly have a long term approach to these investment themes, which we have actually
been covering for years, and try to pick the winners in each industry or subsector
 hence, rebalancings in the AI & Robotics certificate should not happen too often
 however, they should be much more frequent in the Innovation certificate as we will have to
make asset allocation decisions between different subsectors
 Given the high beta nature of tech investments and specifically of AI software names, our
certificates are pretty well diversified with 25-30 stocks on average
 that said, we have a couple of bets in order to generate alpha
 and in order to preserve potential alpha generation, we plan to have no more than 35 stocks
as we consider that extreme diversification does not generally help portfolio performance
 Globally, the Innovation certificate should be less volatile than the AI & Robotics one as it is
exposed to different tech subsectors with specific dynamics
 In a near future, we’d like to launch an AMC… managed by AI
27

OUR INVESTMENT PRODUCTS AND PROCESS
Innovation certificate details
 ISIN: XS0884325415
 Issuer: Natixis
 Management fees: 1.65%
 Performance fees: 15%
AI & Robotics certificate details
 ISIN: XS1035525325
 Issuer: Natixis
 Management fees: 1.65%
 Performance fees: 15%
28

DISCLAIMER
This report has been produced by the organizational unit responsible for investment research (Research unit) of AtonRâ
Partners and sent to you by the company sales representatives.
As an internationally active company, AtonRâ Partners SA may be subject to a number of provisions in drawing up and
distributing its investment research documents. These regulations include the Directives on the Independence of Financial
Research issued by the Swiss Bankers Association.
Financial institutions are accordingly obligated to disclose any conflicts of interest. Special emphasis has been placed on the
following points:
- Has not participated in the issue of securities of the analyzed companies in the last three years.
- Has not been part of a syndicate which has participated in the issue of securities of the analyzed companies in the
last 12 months.
- Holds less than 5% of the voting rights of the analyzed companies.
- Does not have executives who are members of the board of the analyzed company.
Compensation of analysts is not directly related to investment banking transactions.
This document was not submitted to the analyzed companies before publication or distribution.
Although AtonRâ Partners SA believes that the information provided in this document is based on reliable sources, it cannot
assume responsibility for the quality, correctness, timeliness or completeness of the information contained in this report.
The information contained in these publications is exclusively intended for a client base consisting of professionals or qualified
investors. It is sent to you by way of information and cannot be divulged to a third party without the prior consent of AtonRâ
Partners. It cannot be considered under any circumstances as an offer to sell, or a solicitation of any offer to buy financial
instruments.
While all reasonable effort has been made to ensure that the information contained is not untrue or misleading at the time of
publication, no representation is made as to its accuracy or completeness and it should not be relied upon as such. Past and
simulated performances offer no guarantee as to future performances. Any opinions offered herein reflect our current
judgment and may change without notice. AtonRâ Partners cannot be held responsible for the consequences of any decision
made with regard to the information contained in this document.
29

Artificial Intelligence + Robots

More Related Content

What's hot (19)

Similar to Artificial Intelligence + Robots (20)

Artificial Intelligence + Robots