Running head ARTIFICIAL INTELLIGENCE1ARTIFICIAL INTELLIGENCE.docx

Running head: ARTIFICIAL INTELLIGENCE 1
ARTIFICIAL INTELLIGENCE 5
Advantages of Artificial Intelligences, Uploads, and Digital
Minds
First Article
The first article states that the term artificial intelligence
currently impact our lives and civilization. The extents of an
artificial intelligence application are different with extensive
probabilities. In specific, as of improvement and developments
in computer hardware, isolated artificial intelligence algorithm
already exceed the abilities of human specialists today. The
empire of probable uses of artificial intelligence techniques is
vast. Also, this is one of the causes why several corporations
have been deeply investing in AI in current years. Google
Corporation is constructing self-driving automobiles and cars
also has developed ten robotics corporations, Facebook had
released a new investigation facility concentrated on AL
intelligence.
On the other hand, “Apple has industrialized Siri, Microsoft has
constructed Cortana, and Google has developed Deep Mind.” A
UK company who long-term objective is to construct a usual AI
also has previously displayed wide possibility in engaging at
the game of “Go the current world champion.” IBM is investing
a wide amount of assets during applying its “Watson reasoning
computing system” to the health domain, to economics, also to
adapted education. This growth of AI-based services and
systems are getting all bends of the globe. As per the article, AI

is not regarding computing authority. Intelligent machines and
system can also relay on wide amounts of information, to be
used to investigate how to make fine decisions.
This information and data originate from all of the people. Over
many years, Facebook workers have uploaded 250 billion
images. Also they upload around 350 million regularly. The
point that a recent AI artificial intelligence can make more
precise medical detects than doctors might seem astonishing
initially, but it has been accepted that arithmetical implications
are greater to medical judgments through human specialists in
many situations. Progress and development in AI investigation
make it probable to substitute rising amounts of human
occupations with many advanced machines (Sotala, 2012).
Artificial and Intelligences and its Role in
Near Future.
Second Article
The second article describes that AI knowledge has long past
which is constantly and actively growing and modifying. If
emphasis on many intelligent agents, that consists devices that
observe background also reliant on which take many actions to
enhance the chances of success. In the context of a current
digitalized domain, AI is the assets of computer programs,
machines and systems to achieve the creative and intellectual
functions of an individual, self-sufficiently recognize several
methods to resolve any issue, be capable of defining
conclusions also make decisions. Many AI systems can acquire,
which enables people to recover their routine and performance.
The current investigation on AI tools, comprising machine
knowledge, predictive examination, and deep learning
envisioned towards growing the planning, reasoning, learning,
and thinking with action pleasing ability. AI states to the
potential or possibility of computer-controlled robots to perform
functions and tasks that similar to people. In this situation, AI
is used to implement several robots that have several human

rational characteristics, learning and behavior from previous
experience, have human intellectual characteristics, behaviors,
learning from experience, have capabilities to sense, and to
making estimations also describes the meaning of an isolated
situation.
As per the article, robotic technology is widely trending in the
recent life which has attained popularity in several sectors like
industries, schools, hospitals, military, gaming, music,
important science, and several others. AI is a well-organized
means that make software and computers control robotic
discerning with expert arrangements that knowingly exemplify
the intelligent performance, knowledge and efficiently advice
operators. In simple words, AI is generally recognized as the
potential or ability of robotics to take effective decisions and to
resolve several issues. The term artificial intelligence becomes
ver7 beneficial for an organization as it helps to rise the growth
and productivity of an organization (Jahanzaib Shabbir, 2015).
References
Jahanzaib Shabbir, a. T. (2015). Artificial Intelligence and its
Role in Near Future . JOURNAL OF LATEX CLASS FILES, 1-
11.
Sotala, K. (2012). Advantages of Artificial Intelligences,
Uploads, and Digital Minds. International Journal of Machine
Consciousness , 275–291.

JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8,
AUGUST 2015 1
Artificial Intelligence and its Role in Near Future
Jahanzaib Shabbir, and Tarique Anwer
Abstract—AI technology has long history which is actively and
constantly changing and growing. It focuses on intelligent
agents,
which contains devices that perceives environment and based on
which takes actions in order to maximize goal success chances.
In
this paper, we will explain the modern AI basics and various
representative applications of AI. In context of modern
digitalized world,
Artificial Intelligence (AI) is the property of machines,
computer programs and systems to perform the intellectual and
creative functions
of a person, independently find ways to solve problems, be able
to draw conclusions and make decisions. Most artificial
intelligence
systems have the ability to learn, which allows people to
improve their performance over time. The recent research on AI
tools,
including machine learning, deep learning and predictive
analysis intended toward increasing the planning, learning,
reasoning,
thinking and action taking ability [1]. Based on which, the
proposed research intended towards exploring on how the
human intelligence

differs from the artificial intelligence [2]. In addition, on how
and in what way, the current artificial intelligence is clever than
the human
beings. Moreover, we critically analyze what the state-of-the art
AI of today is capable of doing, why it still cannot reach human
level
intelligence and what are the open challenges existing in front
of AI to reach and outperform human level of intelligence.
Furthermore, it
will explore the future predictions for artificial intelligence and
based on which potential solution will be recommended to solve
it within
next decades [2].
F
1 INTRODUCTION
The term intelligence refers to the ability to acquire
and apply different skills and knowledge to solve a given
problem. In addition, intelligence is also concerned with
the use of general mental capability to solve, reason, and
learning various situations [3], [4], [5], [6], [7], [8], [9], [10],
[11], [12], [13], [14], [15], [16], [17], [18], [19], [20], [21],
[22], [23]. Intelligence is integrated with various cognitive
functions such as; language, attention, planning, memory,
perception. The evolution of intelligence can basically is
studied about in the last ten years. Intelligence involves
both Human and Artificial Intelligence . In this case, critical
human intelligence is concerned with solving problems,
reasoning and learning. Furthermore, humans have simple
complex behaviors which they can easily learn in their entire
life [24].
2 WHICH OF THESE AND IN WHAT LEVEL CAN TO-
DAYS ARTIFICIAL INTELLIGENCE DO?

Todays Artificial Intelligence (robotics) has the capabilities
to imitate human intelligence, performing various tasks
that require thinking and learning, solve problems and
make various decisions. Artificial Intelligence software or
programs that are inserted into robots, computers, or other
related systems which them necessary thinking ability [25].
However, much of the current Artificial Intelligence systems
(robotics) are still under debate as they still need more
research on their way of solving tasks. Therefore Artificial
Intelligence machines or systems should be in position to
perform the required tasks by without exercising errors.
In addition, Robotics should be in position to perform
various tasks without any human control or assistance
[24]. Todays artificial intelligence such as robotic cars are
highly progressing with high performance capabilities such
as controlling traffic, minimizing their speed, making from
self-driving cars to the SIRI, the artificial intelligence is
rapidly progressing [26]. The current attention towards por-
traying the artificial intelligence in robots for developing
the human-like characteristics considerably increases the
human dependence towards the technology. In addition,
the artificial intelligence (AI) ability towards effectively
performing every narrower and cognitive task considerably
increases the peoples dependence towards the technology
[27]. Artificial intelligence (AI) tools having the ability
to process huge amounts of data by computers can give
those who control them and analyze all the information.
Today, this considerably increases the threat which makes
someone’s ability to extract and analyze data in a massive
way [24]. Recently, Artificial intelligence is reflected as the
artificial representation of human brain which tries to sim-
ulate their learning process with the aim of mimicking the
human brain power. It is necessary to reassure everyone that
artificial intelligence equal to that of human brain which is
unable to be created [25]. Till now, we operate only part

of our capabilities. As currently, the level of knowledge is
rapidly developing, it takes only a part of the human brain.
As the potential of human brain is incommensurably higher
than we can now imagine and prove. Within human brain,
there are approximately 100 trillion electrically conducting
cells or neurons, which provide an incredible computing
power to perform the tasks rapidly and efficiently. It is
analyzed from the research that till now computer has the
ability to perform the tasks of multiplication of 134,341 by
989,999 in an efficient manner but still unable to perform the
things like the learning and changing the understanding of
world and recognition of human faces [28].
3 WHERE DOES THE HUMAN INTELLIGENCE DIF-
FER FROM AI?
Artificial intelligence refers to the potential of computer
controlled machines/robots towards performing tasks that
that almost or similar to human beings. In this case, Ar-
tificial intelligence is used to develop various robots that
have human intellectual characteristics, behaviors, learning
from past experience, have abilities to sense, and abilities
ar
X
iv
:1
80
4.
01
39
6v
1

[
cs
.A
I]
1
A
pr
2
01
8
AUGUST 2015 2
to making predications and determine meaning of certain
situation [29]. Robotic technology is largely trending in
the current life which has gained popularity in various
sectors such as industries, hospitals, schools, military, music,
gaming, quantum science and many others [28]. Artificial
Intelligence is an efficient means that make computers and
software control robotic thinking with expert systems that
significantly illustrate the intelligent behavior, learning and
effectively advice users. In general, AI is basically known as
the ability or potential of robotics to decide, solve problems
and reason [30]. There are various innovations of Artificial
Intelligence, for example robotic cars which dont require a
driver to control or supervise them. In addition, artificially
intelligent technology (robots) involves smart machines that
process a large amount of data that a human being cant

be in position to perform. By so robotics are assuming
repetitive duties that require creativity and knowledge base.
Furthermore, Artificial Intelligence (AI) is the combination
of various technologies that give chance to robotics to un-
derstand, learn, perceive or complete human activities on
their own [2]. In this case, Artificial Intelligence programs
(robots) are built for a specific purpose such as learning,
acting and understating whereas humans intelligence is
basically concerned with various abilities of multitasking.
In general, an Artificial Intelligence tool is majorly con-
cerned with emphasizing robotics which portrays human
behaviors. But however, Artificial Intelligence may fail out
at some points due to differences in human brain and
computers. In brief, Artificial Intelligence has the potential
to mimic human character or behaviors [31]. Furthermore,
Artificial intelligence is currently partially developed with-
out advanced abilities to learn on their own but instead
given commands to act on. This will be the ultimate future
of artificial intelligence, where the AI machines will be
recognized the human behavior and emotions and will train
their kernel as per it [32].
4 WHY CANT WE TELL THAT TODAYS AI IS AS
CLEVER AS HUMAN BEINGS?
Generally, there are various paths towards building the
intelligent machines that enables the humans to build the
super-intelligent machines and provide ability to machines
towards redesigning their own programming in order to
increase their intelligence level, which is usually considered
as the intelligence explosion. In contrast, the shielded hu-
man hunt is basically the emotion. The breakthrough of
AI technology can frighten the humanity in a way that
machine are unable to effectively transmit the emotions.
So, there may be possibility that AI can support us with
the tasks and functions which usually not involves the

feelings and emotion. Till now, AI machines are not able
to control their process, for which they need the intelligence
and mind of human beings [32]. But AI development with
same pace may cause threat to the humanity, because the
self-learning ability may cause the AI machines to learn
destructive things, which may cause killing of humanity in
a drastic way. In general, there exist various characteristics
which distinguish human level intelligence with Artificial
intelligence and they include the following;
Thinking ability; it can be both positive as well as neg-
ative because of having emotions, which are not with AI
machines. The lack of machines emotion may lead to de-
structive in a situation where emotions are required. Russel
Stuart believes that machines would be able to think in a
weak manner. In general, there are things that computers
cannot do, regardless of how they are programmed and
certain ways of designing intelligent programs are doomed
to failure sooner or later [33]. Therefore, most accurate
idea would be to think that it is never going to make the
machines have a thought at least similar to human The lack
of thinking ability of machines may cause lack in passing
the behavioral test. What was later called the Turing Test,
proposed that a machine be able to converse before an
interrogation for five minutes for the year 2000 and in fact, it
was partly achieved. It is concluded then, that the machines
can actually think, although they can never have a sense
of humor, fall in love, learn from experience, know how
to distinguish the good from the bad and other attitudes
of the human. Artificial Intelligence: A Modern Approach
dedicates its last chapter to wonder what would happen
if machines capable of thinking were conceived [34]. Then
it is when we ask ourselves if it is convenient to continue
with this project, take risks and follow an unknown path
and believe that what may happen will not be negative.
Russel and Norvig believe that AI machines role are to be

more optimistic. They believe that intelligent machines are
capable of ”improving the material circumstances in which
human life unfolds” and that in no way can they affect
our quality of life negatively. Reasoning; algorithms mimic
the phased reasoning that people use to solve puzzles and
guide logical conclusions. For complex tasks, algorithms
may require huge computational resources; most of them
experience ’combinatorial explosions’. Memory or required
computer time will be astronomical with tasks of a certain
size. Finding a more effective algorithm to solve the problem
is a top priority. Humans usually use quick and intuitive
judgment rather than gradual deductions modeled by ear-
lier AI studies. I made progress using the ”sub-symbol”
solution to the problem. The materialized approach of the
agent highlights the importance of sensory - motor skills for
higher reasoning. The search of the neural network attempts
to imitate the structure in the brain that generates this
skill. The statistical approach to AI mimics human guessing
abilities. Planning ability; the objection of those who defend
human intelligence against that of machines is based on the
fact that they do not possess creativity or consciousness.
Planning and creativity could be defined as the ability to
combine the elements at our disposal to give an efficient, or
beautiful, or shrewd solution to a problem we are facing.
That is, we call creativity what we have not yet been able
to explain and reproduce mechanically in our behavior.
However, the functioning of artificial neural networks can
also be considered creative but little predictable and plan
able [35]. Action taking ability; the action taking ability of
humans are based on emotions, deep thinking and its com-
parison with how much it is beneficial for the human beings
while AI only takes actions either based on their coding,
therefore, the lack of emotions and comparison about good
and bad increases threats. AI machines are actually very
intellectually limited, although they may become brilliant

AUGUST 2015 3
in specialization. A child in his earliest childhood is able to
learn to put the triangle in the hollow of the triangle of the
toy, can recognize the sounds of animals and begin to apply
what he learned on a stage in different ones. Machines are
not able to do this at the same time without being previously
trained for each task. Generally, there is no neural network
in the world capable of identifying objects, images, sounds
and playing video games at the same time as people [36]. Its
limitations are obvious even within the same field of action:
when Google’s Deep Mind created a system to pass Atari
video games, relatively simple for a computing entity due
to its two-dimensional progress based on reflexes and trial
and error, its neural networks they had to be trained every
time a game was completed. Machines are not capable of
transferring what they learn to other scenarios that catch
them off guard as well as humans can adapt, make use
of logic, creativity, ingenuity and reason in any situation,
however strange and new they may be [37].
Knowledge representation; it includes the problems ma-
chines which expresses the relationships between objects,
properties, categories, objects, situations, events, states, and
times. The potential cause and effect of knowledge repre-
sentation is based on what we know about what others
know of many other well-studied domains. The concept of
”what is present” is an ontology that officially describes
a set of objects, relationships, concepts, and properties so
that the softwares agent can significantly interpret it. The
semantics of this data are recognized as logic to describe
roles and descriptions which usually realized as ontology
web language classes, properties, and individuals. The most

common ontology is called a top ontology that provides
the basis of all other knowledge and acts as an intermedi-
ary between domain ontology covering specific knowledge
about a specific knowledge field. Such formal knowledge
representation is based on knowledge using content-based
indexing and searching, interpretation of scenarios, support
for clinical decisions, automatic reasoning [38]. Perception;
the machine perception is the ability to derive aspects of
the world using input signals from cameras, microphones,
sensors, sonar etc., while the computer vision is the ability
to analyze visual input. Learning ability; the development of
new algorithm considerably helps the AI machines to learn
to write like humans and is able to recognize and draw
very simple visual concepts. Generally, the main virtue of
human beings is their speed and diversity, when it comes
to learning new concepts and applying them in new situa-
tions. Computers usually have hard time generalizing from
particular samples. In addition, to assessing the ability of the
program to learn concepts, they asked people to reproduce
a series of characters that had also been plotted by the ma-
chine [38]. Then, they compared them and asked different
people what they thought had been done by humans and
what by a program, in an adaptation of the classic Turing
test that the researchers call visual Turing test. Basically the
Turing test is that someone in a room is asking questions to
determine if the answers and interactions they receive come
from a person or a machine. Therefore, machine learning
ability machines significantly makes it much better than that
of human being [38]. The natural language process system;
they give machines the ability to read and understand
human language. A sufficiently powerful natural language
processing system enables the use of user interfaces in
natural language and gains direct knowledge from written
sources such as news texts. Some simple natural language
processing applications include information retrieval, text

mining, machine translation etc [39]. The general way to
extract and process the natural language, it is essential to use
semantic indexing. These indices are expected to increase
efficiency as the user speeds up the processor and lower the
cost of storing the data despite the user’s input being large
[27].
5 GIVE A COMPREHENSIVE SURVEY OF WHAT CAN
TODAYS AI SUBFILELDS DO?
Todays artificial intelligence is creeping into our daily lives
by using the GPS navigation and check-scanning machines.
The use of artificial intelligence (AI) in business contributes
to the potentialization of various areas of daily life such
as customer service, finance, sales and marketing, admin-
istration and technical processes in various sectors. Un-
doubtedly, over the next few years, digital efforts will no
longer be isolated projects or initiatives in companies, but
the adoption of technologies such as AI at all levels and
processes of companies will be a reality to increase their
competitiveness. AI begins to integrate into the activities
of business. It is important to consider that it has not
arrived to replace human tasks, but to complement them
and allow people to develop their potential and creativity to
the maximum. Introduction of new technologies is a tool for
the prevention and fight against corruption, the traceability
of electronic actions, and the security that surrounds their
management favors confidence in management. This essay
will prove how artificial intelligence can improve efficiency
of people, help create jobs, and begin to make our soci-
ety safer for our children. Currently, massive research on
artificial intelligence significantly improving the world, in
which majority of the tasks was performed by the machines,
while the role of humans will be to control them. This leads
towards the question that is artificial intelligence exceeds the
performance level of humans and provide human task in an
efficient, quicker and economical level? The paper is about

the present and future of artificial intelligence technology
and compares it with human intelligence [39]. The role of
this paper is to explore where we are today and what will
be the ultimate future of AI technology, if we continuously
applied in every field. Further, we will analyses the potential
AI techniques and their potential benefits for improving the
system.
6 BRIEFLY EXPLAIN TECHNICAL BACKGROUND AS
WELL?
Artificial Intelligence has facilitated us in almost every field
of life and has immense scope in future for more produc-
tivity and betterment. The origin of artificial intelligence
goes back to the advances made by Alan Turing during
World War II in the decoding of messages. The term as
such was first used in 1950, but it was only in the 1980s
when research began to grow with the resolution of algebra
equations and analysis of texts in different languages. The
definitive takeoff of Artificial intelligence has come in the
AUGUST 2015 4
Fig. 1: Abilities of Artificial Intelligence
last decade with the growth of the internet and the power
of microprocessors [40]. ”Artificial intelligence may be the
most disturbing technology the world has ever seen since
the industrial revolution” Paul Daugherty, Accenture’s chief
technology officer, recently wrote in an article published
by the World Economic Forum. This field is now boom-
ing due to the increase in ubiquitous computing, low-
cost cloud services, new algorithms and other innovations,
adds Daugherty. Developments in Artificial Intelligence go

hand in hand with the development of processors that
over time have made them start to see these technologies
as intellectual, even changing our idea of intellect and
forthcoming the perceptions of ’machine’, traditionally un-
intelligent capacity previously assigned exclusively to man.
The AI was introduced to the scientific community in 1950
by the English Alan Turing in his article ”Computational
Machinery and Intelligence.” Although research on the de-
sign and capabilities of computers began some time ago, it
was not until Turing’s article appeared that the idea of an
intelligent machine captured the attention of scientists. The
work of Turing, who died prematurely, was continued in
the United States by John Von Neumann during the 1950s
[41]. His central contribution was the idea that computers
should be designed using the human brain as a model. Von
Neumann was the first to anthropomorphize the language
and conception of computing when speaking of memory,
sensors etc. of computers. He built a series of machines
using what in the early fifties was known about the human
brain, and designed the first programs stored in the memory
of a computer [27].
McCulloch (1950) formulate a radically different posi-
tion by arguing that the laws governing thought must be
sought between the rules that govern information and not
between those that govern matter. This idea opened great
possibilities for AI. In addition, Minsky (1959) modified his
position and argued that imitation of the brain at the cellular
level should be abandoned. The basic presuppositions of
the theoretical core of the AI were emphasis on recognition
of thought that can occur outside the brain [?], [42], [43],
[44]. On 1958, Shaw and Simon design the first intelligent
program based on their information processing model. This
model of Newell, Shaw and Simon was soon to become
the dominant theory in cognitive psychology. At the end

of the 19th century, sufficiently powerful formal logics were
obtained and by the middle of the 20th century, machines
capable of make use of such logics and solution algorithms.
7 WHAT IS MISSING TO TODAYS AI STILL TO BE
CALLED HUMAN LEVEL INTELLIGENCE?
Humans are different from Artificial Intelligence machine
physically in a sense that human race usually experiences
the same physical features while the machines takes several
forms and shapes. The trans-humanist vision analysis ex-
hibits us to believe that brains are principally the computers.
AI reports are the silicon based machines, which was con-
trolled by using the algorithm that reinforces entire internet
business. AI believes that once the computers have adequate
advanced algorithms, then they will be capable to replicate
and enhance the human mind [45]. The tests which exhibits
how much AI is distinct from the human intellectual are as
follows Turing test; in order to evaluate on what intelligence
means and on how the machine intelligence is different than
the human intelligence, the Turing test strongly provide the
AUGUST 2015 5
Fig. 2: Timeline evolution of Artificial Intelligence
Fig. 3: Turing test
essential insights to the AI field, which emphasis on how
the machine simulates the human thinking. The algorithmic
aspects of AI tools should pass the Turning Test [46]. This
algorithm will not essentially result in the AGI but may also
lean towards applied artificial intelligence. The algorithm
tuned through Turing process can also significantly define

and passed it.
Eugene Goostman Test; Goostman tests the Turing test
and concluded that it has 33% fooling, which intended him
to propose test that should rely on AI in order to efficiently
solve the particular tasks that was quite near to AGI. He
also concluded that it is not possible for AI machines to be
much efficient than that of human because their always acts
human actors to work the AI machine.
8 IMPORTANCE OF ARTIFICIAL INTELLIGENCE
Artificial Intelligence will revolutionize the way in which
different companies across compete and grow across the
world by representing a new production factor that can
drive business profitability [4], [5], [6], [47], [48], [49]. In
order to realize the opportunity of AI, most the companies
in the world are already developing actively in various
Artificial Intelligence strategies [47].In addition, they should
focus on developing responsible AI systems aligned with
ethical and moral values that lead to positive feedback and
empower people to do what they know best such as innova-
tion [50]. With the introduction of successfully implemented
Artificial Intelligence (AI) solutions, many industries across
the earth can benefit from increased profitability and still
count on economic growth. To capitalize on this opportu-
nity, the study identifies eight strategies for the successful
implementation of AI that focuses on adopting a human-
centric approach and taking innovative and responsible
measures for the application of technology to companies
and organizations in the world [51], [52], [53], [54], [55]. The
construction of intelligent machines in various industries
presupposes the existence of symbolic structures, the ability
of them to demand and the existence of knowledge (raw
material). Once artificial intelligence has intelligence equal
to or greater than man’s, political and social change will

inevitably arise, in which AI has all the advantages of
gaining if it realizes that it does not need humans to colonize
the universe [39], [56], [57], [58]. Recent advancement in
artificial technology depicts orbiting communications satel-
lites in the space with its 486 processors. In the future, self-
replicating artificial intelligence could easily be made with
all human colonies outside the earth, and the human race
will never be able to fight in the empty space on equal terms
[45].
9 KNOWLEDGE BASED TOWARDS UNDERSTAND-
ING THE NATURAL LANGUAGE
The most significant characteristics of the natural language
are that it has the ability to serve as its particular meta-
language. It is easily possible to utilize the natural language
AUGUST 2015 6
in order to provide instruction in the use as well as describe
the language itself (Weston et al., 2015). It is because, human
usually capable to utilize their natural language to describe
about natural language itself. The advancement in the nat-
ural language understanding significantly provide effective
educable cognitive agent role whose task domain contains
the understanding of language and whose discourse do-
main contains the knowledge of their own language.
10 SPATIAL FRAMES OF REFERENCE COMPUTA-
TION FOR NARRATIVE TEXT UNDERSTANDING
This AI technique provide the spatial references which
explicitly build in order to indicate on how the reference

frame should be established and provide the determination
of burden of its determination is left to the hearers or the
readers inferences. Its also intended towards developing
heuristics which significantly aid for dynamically comput-
ing of reference frames. The rule based expert systems; they
have important characteristics related to multiple produc-
tion rules and levels that embody the knowledge of the
distinct document images characteristics. Its includes the
inference engine which utilizes the knowledge base in order
to have identities interpretations of several blocks logical
within the image [45], [59], [60]. The control inference engine
is applied in several distinct rules levels on the image data
for blocks identification. The lowest rule level is the knowl-
edge rule that examines the intrinsic blocks properties. The
control rules act and guide the search as the mechanism of
focuses on attention. The strategy rules determine whether
the constant image interpretation has been achieved or not.
The cognitive letter recognition model; it is based on the
feature models utilization. Within the acquisition time, the
letters is subjected towards developing the object oriented
quad tree model. In general, all of the acquired letters are
integrated into object knowledge, similar quad tree and
density values. At the time of recognition, the densities and
quad tree analysis of the current letter are related to store
quad tree density.
11 WHAT ARE OPEN CHALLENGES?
Challenges of AI technology include the following; Within
near future, the artificial intelligence goals were to affect
the society from law and economics in several technical
terms including security, verification, validity, and control
[45], [61], [62], [63], [64], [65]. The major short-term threats of
Artificial Intelligence include the devastating race of arms in
fatal autonomous weapons and full dependence of our life
on technology will eventually lead to unemployment prob-

lem, social discrimination and power inequality in societies.
long term use of robotics will create a big challenge includes
to human beings as they will take over the world. In addi-
tion, with in a given period of time, AI will become better in
solving tasks compared to human beings hence lose of jobs.
For example drivers will be ruled over by robotic cars, tellers
will be out competed by robotic tellers and many others [66],
[67], [68]. As a result of Artificial Intelligence out competing
humans will create a big challenge on human thinking.
But in the positive side, the AI technology advancement
can also considerably aid for eradicating the disease, war
and poverty level. Rapid research of Artificial intelligence
in robotics is reflected as the large existential threats that
are faced by the humanity. The destructive method de-
veloped in the artificial intelligent robot can also increase
the risk for the super-intelligent system. For example, the
ambitious project of geo-engineering can wreak havoc of
ecosystem. This considerably increases the concerns about
the AI system advancement which is not malevolence but
have increased competency. Although, the super-intelligent
artificial intelligence tools can considerably aid in fulfilling
the goals un-alignment of goals may cause an increasing
problem [26], [31], [36], [37], [50], [69], [70], [71], [72], [73].
Artificial intelligence is simply the next wave of automation,
which considerably allows the machines to do tasks that
previously required attention and human intelligence. In the
short term it can replace people, but above all, it changes
the nature of the work that humans do. In the long run,
automation creates more and different types of jobs, which
is why not everyone has jobs today. Further, challenges
include the large dependence of people towards technol-
ogy for their work. This considerably increases several
psychological, physical and mental issues. In order hand,
it considerably reduces the unemployment, economic and
power balance and disability. It is believed that until the

end of 2021, there will be the beginning of a disruptive
tidal wave of artificially intelligent robots in our daily life,
such solutions powered by cognitive or artificial intelligence
technology will significantly displace the jobs, with the great
impact felt in logistics, transportation, consumer services
and educational process. Furthermore, it can considerably
increase the privacy, security and authenticity issues within
the society.
12 WHAT ARE SOLUTION SUGGESTIONS?
To reduce the destructive effects of AI, it is essential to
develop the symbolic approach, which should allow us to
operate with weakly formalized representations and their
meanings. The success and effectiveness of solving new
problems depend on the ability to allocate only essential
information, which requires flexibility in the methods of
abstraction. While a normal program sets one’s own way of
interpreting the data, which makes its work look prejudiced
and purely mechanical. The intellectual task, in this case,
solves only the person, the analyst or the programmer,
not knowing to entrust this machine. As a result, a sin-
gle abstraction model is created, a system of constructive
essences and algorithms. And flexibility and universality
translate into significant resource costs for non-typical tasks,
that is, the system from the intellect returns to brute force.
Furthermore, the hybrid approach should also be developed
in order to provide the synergistic combination of neural
and symbolic models achieves a full range of cognitive
and computational capabilities. For example, expert rules
of inference can be generated by neural networks, and
generating rules are obtained through statistical training.
Supporters of this approach believe that hybrid information
systems will be much stronger than the sum of different
concepts separately. Furthermore, to decrease the potential
threats of artificial intelligent technology, the rational risk
management process that includes the potential principles

AUGUST 2015 7
Fig. 4: Rule-based expert system structure
of adopting the expensive precautions, which have high cost
even for the lower probability risks, can considerably pro-
vide the benefits (Russell et al., 1995). The global nature of
artificial intelligence risks if fails to transfer the ethical goals
then its absolutely reasonable to estimate the longer-term AI
research risks as even larger than those of climate change.
It is also essential to adopt effective information system
which should be established in order to effectively improve
the artificial intelligence safety by initiating the awareness
on the experts working part on AI, decision-makers, and
investors. The risks information integrated with AI progress
should also have to understandable and accessible for the
broad audience. In addition, it is also instigated to adopt
the AI safety tools in order to reduce the authentication and
security issues. It is also essential to develop an effective
global coordination and cooperation system in order to
build the competitive environment in which the dangerous
race of artificial intelligent arms should be reduced [?],
[46], [51], [74], [75], [76]. It is believed that although the
future of artificial intelligence has a tremendously beneficial
impact on the economy and daily lives of Americans still it
increases the issues related to privacy, security, unemploy-
ment, technological dependency, and authenticity within the
society. Therefore, it is essential to develop the preventive
control solution for mitigating or reducing such challenges.
13 WHAT ARE FUTURE PREDICTIONS FOR AI?
There is a great increase in the discussion about the impor-

tance of AI in the recent time leading to future discussions
about the existence of Artificial Intelligence in the world.
The idea of creating AI is aimed at making human life easier.
However, there is still a big debate about advantages and
disadvantages of AI in the whole [77], [78], [79], [80], [81],
[82], [83]. With the introduction and successful implementa-
tion of Artificial Intelligence (AI) solutions, many industries
in the world are and will benefit from increased profitability
and will still have good economic growth rates. In addi-
tion, artificial Intelligence opportunities will be aiming at
innovative, human centered approaches and measuring the
applicability of robotic technology to various industries and
companies in the entire world. Artificial Intelligence will
also revolutionize the way different companies in the world
grow and compete by representing new production ideas
that will derive profitability in businesses [77], [84], [85],
[86], [87]. So as to realize such opportunities, it will require
most of the companies in the world to become more active
in the development of various Artificial Intelligence strate-
gies such as placing human factors to central nucleus. In
addition, they will focus on developing various responsible
Artificial Intelligence machines having moral and ethical
values which will result into positive results and empow-
erment of people to do things that they are well versed
with. Construction of various Artificial Intelligence systems
will help the entire world to industrial sector to presuppose
the available symbolic structures such as, the ability to
reason and also knowledge existence. In addition, at the
time Artificial Intelligence acquires intelligence greater or
equal to that of human beings, there will be a concern
about social and political change .In furthermore, AI will
have all the advantages of colonize the world without the
help of human beings. In the near future, self-replicating AI
could be made where human colonies beyond the earth will
never have potentials to fight in the free space with critical

terms. The future Artificial Intelligence in various regions
in the world may be as a result of various investigation
technologies such as stellar travel, teleportation and others.
AUGUST 2015 8
14 WHICH LEVELS WILL IT REACH AND WHICH
ISSUES WILL BE SOLVED IN NEXT DECADES?
In this case, it is more likely that Artificial Intelligence
innovations will strongly emerge in conceivable future. In
next decades, the future of AI will be concerned on improv-
ing speech, voice, video conferencing and face recognition.
Further, Artificial Intelligence will aid for providing the
personal assistances and fully automate systems, which will
provide assistance in monitoring and surveillance, perform-
ing heavy workloads and many others [88], [89], [90], [91].
In addition, the future of Artificial Intelligence technology
such as robotics will be ensuring self-driven cars, delivery
robots and many others. With the great improvement in
computer versions and legged locomotion, the robots within
environments will become more practical hence helping
in agriculture and other service settings. In addition, the
robotics will improve on the service delivery hence reducing
domestic chores [11], [12], [13], [14], [15], [16], [88]. Further-
more, as Artificial Intelligence robotics is developing search
engines, there will be provision of personal assistance and
language gasps by use of mobile devices. The development
of search engines will lead to significant synthesize and
improvement of the quality of information. Furthermore,
Artificial Intelligence tool will improve the medical and bio-
logical system hence reducing the complexity and volume of
information challenges concerned human abilities. Artificial

intelligence will be used in the algorithm that materializes
various systems and programs. Artificial intelligence will
consist of specific hardware and software which intends to
imitate the way of human brain performance. In general, the
areas of Artificial Intelligence application will widely cover
both emerging and traditional technologies. Recent research
on AI considerably provide the potential effects on organi-
zation and industries, for example, the AI technology will
be aiming at improving the area of data science to almost
9.6%, business intelligence to 7.8%, patient and health care
to about 6.3%, speech recognition 5.3%, computer vision
5.6%, improve defense and aerospace system to about 5.3%
and natural language processing to about 5.1%. The role
of Artificial Intelligence tools on product manufacturing
operations will lead to toward employment, flexibility and
responsive chain of supply. In addition, AI roles will also
result to reliable forecasting demands, inventory accuracy
and optimization of schedules. The roles of AI will therefore
benefit quicker, smarter and environmental efficient process.
AI application in security and defense will majorly focus
on infrastructure protection. Currently Artificial Intelligence
facilitates the power plant, airport and economic sectors
which are quite hard to detect attacks, individuals anoma-
lous predication of disruption by man-made and natural
causes [92].
Within logistics area, the intervention of Artificial Intel-
ligence will contain efficient vehicles that will be in position
to route and make necessary adaptive delivery schedules. In
the financial service sector, Artificial Intelligence tools will
contain system failure and risk alerts aiming at decreasing
malicious attacks various financial systems; such as fraud,
market manipulation and reduction in market volatility
and trading costs. In the agriculture sector, the intelligent
solutions will provide intelligent production mechanisms

for processing, consumption, storage and distribution [92].
The Artificial solution will also provide given timely data
on crops that will involve use of proper materials such
as chemicals and fertilizers. Artificial Intelligence will also
be used in consumer goods and services so as to utilize
machine learning processes to match consumer demand
and enable them get best ;practices at reduced ;prices.
In communication sector, there will be improvement in
bandwidth and storage and Web translation languages.
Within the education sector, Artificial Intelligence solution
will intervene basic meaningful adaptive learning basing on
adaptive learning complemented by individual learning in
the classroom, accurate measurement of student’s sensitivity
and development of students. Easy manual techniques and
judgment will be supplemented by artificial intelligence.
The medical & health-care will provide various health eval-
uations to patients, decision support for prescribing drugs
and indication. Artificial intelligence will be used in large-
scale genome researches to determine new drugs, give nec-
essary support for finding new genetic problems, efficiency
and safety. The evidence based health and medicine will
help various physicians gain confidence which will require
supplementary support by patients [24]. In customer service
sector, AI systems will provide the virtual assistants which
will aim at increasing the reproduction and interpretation
abilities of human language with greater precision. For
example, chat bots redraw the landscape of the IT ecosys-
tem. They will replace themselves and applications, and
service personnel in companies, and even entire operating
systems. Chat-bot (Chat-bot) - this program will contain an
interlocutor, which will be designed to communicate and
help people. At the other end, there will be a complex
system based on several Artificial Intelligence technologies.
Chat bots, oriented to business tasks, will help to can take
up best flights, diet, fitness trainings, booking of a hotel,
make purchase, that is to say; they will represent a unique

sub-sector of assistance and advice. Personal assistants are a
kind of incarnation of chat bots, although more common
because the technology will be developed by the largest
IT companies [10], [11], [12], [13], [14], [15], [71]. Currently,
hundreds of millions of people interact with personal digital
assistants on platforms such as Google, Apple, Amazon,
Facebook and others. This technology with the help of
personal assistants and chat-bots will be more effective
which will make a great transition from the graphical user
interface (GUI) to the Conversational User Interface (CUI)
the key trend of the next decades. The predictive algorithms
and machine learning based on AI tools will provide sales
forecasts in specific markets and also effective provision
and optimization of inventory as they will help forecast in-
come and determine the necessary quantities of a particular
input. In addition, modern Artificial Intelligence systems
will control robotics to provide surveillance, security and
attacks without threatening the human life in Warfield. AI
applications in robotics will have diverse objectives related
to automation of military applications, industrial processes
and space exploration. The use of AI technology in medicine
and surgeries will significantly help in provision of safe
work as the machinery occupied will reduce the degree of
error that could occur in surgery, avoid a tragic outcome
[93]. The disaster recovery and management application of
AUGUST 2015 9
Fig. 5: Application areas of AI
Artificial Intelligence will considerably remark the provision
of remedial and control actions in the aftermath of man-
made and environmental disasters. Within disasters, the

considerable will optimize the mobile networks and allocate
the smart bandwidth [94]. Further, the satellite feed and
unmanned drones having image recognition and processing
feature will help in assessing damage of infrastructure and
provide predictions aimed at avoiding traffic congestion and
various structural stability by adopting the adaptive routing
system.
15 CONCLUSION
In this way, artificial intelligence can achieve great discov-
eries and advances for humanity due to its multiple possi-
bilities. Most artificial intelligence systems have the ability
to learn, which allows people to improve their performance
over time. The adoption of AI outside the technology sector
is at an early or experimental stage [?], [29], [34], [43], [44],
[44], [48], [49], [54], [55], [59], [60], [70], [72], [72], [95],
[95],
[96], [97], [98], [99], [100]. The evidence suggests that AI
can provide real value to our lives.AI bases its operation
on accessing huge amounts of information, processing it,
analyzing it and, according to its operation algorithms,
executing tasks to solve certain problems. Due to the new
computing architectures of the cloud, this technology be-
comes more affordable for any organization.
ACKNOWLEDGMENTS
The authors would like to thank...
REFERENCES
[1] J. Shabbir and T. Anwer, “A Survey of Deep Learning
Techniques
for Mobile Robot Applications,” ArXiv e-prints, Mar. 2018.
[2] U. Neisser, G. Boodoo, T. J. Bouchard Jr, A. W. Boykin, N.
Brody,

S. J. Ceci, D. F. Halpern, J. C. Loehlin, R. Perloff, R. J.
Sternberg
et al., “Intelligence: Knowns and unknowns.” American
psycholo-
gist, vol. 51, no. 2, p. 77, 1996.
[3] R. Feuerstein, The Dynamic Assessment of Cognitive
Modifiability:
The Learning Propensity Assessment Device : Theory,
Instruments
and Techniques. ICELP Press, 2002. [Online]. Available:
https://books.google.com.pk/books?id=-3vsAAAAMAAJ
[4] M. Milford, C. Shen, S. Lowry, N. Suenderhauf, S. Shirazi,
G. Lin,
F. Liu, E. Pepperell, C. Lerma, B. Upcroft et al., “Sequence
searching with deep-learnt depth for condition-and viewpoint-
invariant route-based place recognition,” in Proceedings of the
IEEE Conference on Computer Vision and Pattern Recognition
Work-
shops, 2015, pp. 18–25.
[5] K. Fragkiadaki, S. Levine, P. Felsen, and J. Malik,
“Recurrent net-
work models for human dynamics,” in Computer Vision (ICCV),
2015 IEEE International Conference on. IEEE, 2015, pp. 4346–
4354.
[6] S. Niekum, S. Osentoski, G. Konidaris, S. Chitta, B. Marthi,
and
A. G. Barto, “Learning grounded finite-state representations
from
unstructured demonstrations,” The International Journal of
Robotics
Research, vol. 34, no. 2, pp. 131–157, 2015.

[7] C. Devin, A. Gupta, T. Darrell, P. Abbeel, and S. Levine,
“Learn-
ing modular neural network policies for multi-task and multi-
robot transfer,” in Robotics and Automation (ICRA), 2017 IEEE
International Conference on. IEEE, 2017, pp. 2169–2176.
[8] C. Finn, X. Y. Tan, Y. Duan, T. Darrell, S. Levine, and P.
Abbeel,
“Deep spatial autoencoders for visuomotor learning,” in
Robotics
and Automation (ICRA), 2016 IEEE International Conference
on.
IEEE, 2016, pp. 512–519.
[9] A. A. Rusu, M. Vecerik, T. Rothörl, N. Heess, R. Pascanu,
and
R. Hadsell, “Sim-to-real robot learning from pixels with
progres-
sive nets,” arXiv preprint arXiv:1610.04286, 2016.
[10] S. Mohamed and D. J. Rezende, “Variational information
max-
imisation for intrinsically motivated reinforcement learning,” in
Advances in neural information processing systems, 2015, pp.
2125–
2133.
[11] Y. Zhu, R. Mottaghi, E. Kolve, J. J. Lim, A. Gupta, L. Fei-
Fei,
and A. Farhadi, “Target-driven visual navigation in indoor
scenes
using deep reinforcement learning,” in Robotics and Automation
(ICRA), 2017 IEEE International Conference on. IEEE, 2017,
pp.
3357–3364.

[12] F. Cruz, J. Twiefel, S. Magg, C. Weber, and S. Wermter,
“In-
teractive reinforcement learning through speech guidance in a
domestic scenario,” in Neural Networks (IJCNN), 2015
International
Joint Conference on. IEEE, 2015, pp. 1–8.
[13] A. Vinciarelli, A. Esposito, E. André, F. Bonin, M.
Chetouani, J. F.
Cohn, M. Cristani, F. Fuhrmann, E. Gilmartin, Z. Hammal et al.,
“Open challenges in modelling, analysis and synthesis of human
behaviour in human–human and human–machine interactions,”
Cognitive Computation, vol. 7, no. 4, pp. 397–413, 2015.
[14] J. Doshi, Z. Kira, and A. Wagner, “From deep learning to
episodic
memories: Creating categories of visual experiences,” in
Proceed-
https://books.google.com.pk/books?id=-3vsAAAAMAAJ
AUGUST 2015 10
ings of the third annual conference on advances in cognitive
systems
ACS, 2015, p. 15.
[15] X. Wang, D. Fouhey, and A. Gupta, “Designing deep
networks for
surface normal estimation,” in Proceedings of the IEEE
Conference
on Computer Vision and Pattern Recognition, 2015, pp. 539–
547.

[16] H. Cuayáhuitl, S. Keizer, and O. Lemon, “Strategic
dialogue
management via deep reinforcement learning,” arXiv preprint
arXiv:1511.08099, 2015.
[17] B. M. Lake, T. D. Ullman, J. B. Tenenbaum, and S. J.
Gershman,
“Building machines that learn and think like people,”
Behavioral
and Brain Sciences, vol. 40, 2017.
[18] E. Ohn-Bar and M. M. Trivedi, “Looking at humans in the
age
of self-driving and highly automated vehicles,” IEEE
Transactions
on Intelligent Vehicles, vol. 1, no. 1, pp. 90–104, 2016.
[19] J. Wei, H. Liu, G. Yan, and F. Sun, “Robotic grasping
recognition
using multi-modal deep extreme learning machine,”
Multidimen-
sional Systems and Signal Processing, vol. 28, no. 3, pp. 817–
833,
2017.
[20] M. Mathieu, C. Couprie, and Y. LeCun, “Deep multi-scale
video prediction beyond mean square error,” arXiv preprint
arXiv:1511.05440, 2015.
[21] G. Chen, D. Clarke, M. Giuliani, A. Gaschler, and A.
Knoll,
“Combining unsupervised learning and discrimination for 3d
action recognition,” Signal Processing, vol. 110, pp. 67–81,
2015.
[22] J. Wulff and M. J. Black, “Efficient sparse-to-dense optical

flow
estimation using a learned basis and layers,” in Proceedings of
the
IEEE Conference on Computer Vision and Pattern Recognition,
2015,
pp. 120–130.
[23] J.-R. Ruiz-Sarmiento, C. Galindo, and J. Gonzalez-
Jimenez,
“Scene object recognition for mobile robots through semantic
knowledge and probabilistic graphical models,” Expert Systems
with Applications, vol. 42, no. 22, pp. 8805–8816, 2015.
[24] L. S. Gottfredson, “The general intelligence factor,” 1998.
[25] Y. Zhang, D. K. Robinson, A. L. Porter, D. Zhu, G. Zhang,
and J. Lu, “Technology roadmapping for competitive technical
intelligence,” Technological Forecasting and Social Change,
vol. 110,
pp. 175–186, 2016.
[26] M. Turan, Y. Almalioglu, H. Araujo, E. Konukoglu, and M.
Sitti,
“A non-rigid map fusion-based direct slam method for endo-
scopic capsule robots,” International journal of intelligent
robotics
and applications, vol. 1, no. 4, pp. 399–409, 2017.
[27] D. C. Parkes and M. P. Wellman, “Economic reasoning and
artificial intelligence,” Science, vol. 349, no. 6245, pp. 267–
272,
2015.
[28] T. Y. Wong and N. M. Bressler, “Artificial intelligence
with deep
learning technology looks into diabetic retinopathy screening,”

Jama, vol. 316, no. 22, pp. 2366–2367, 2016.
[29] M. Turan, Y. Almalioglu, E. Konukoglu, and M. Sitti, “A
deep
learning based 6 degree-of-freedom localization method for en-
doscopic capsule robots,” arXiv preprint arXiv:1705.05435,
2017.
[30] M. Turan, Y. Almalioglu, H. Gilbert, A. E. Sari, U. Soylu,
and M. Sitti, “Endo-vmfusenet: deep visual-magnetic sensor fu-
sion approach for uncalibrated, unsynchronized and asymmet-
ric endoscopic capsule robot localization data,” arXiv preprint
arXiv:1709.06041, 2017.
[31] M. Turan, J. Shabbir, H. Araujo, E. Konukoglu, and M.
Sitti, “A
deep learning based fusion of rgb camera information and mag-
netic localization information for endoscopic capsule robots,”
International journal of intelligent robotics and applications,
vol. 1,
no. 4, pp. 442–450, 2017.
[32] D. M. Martı́nez and J. C. Fernández-Rodrı́guez, “Artificial
intel-
ligence applied to project success: a literature review,” IJIMAI,
vol. 3, no. 5, pp. 77–84, 2015.
[33] J. Weston, A. Bordes, S. Chopra, A. M. Rush, B. van
Merriënboer,
A. Joulin, and T. Mikolov, “Towards ai-complete question
answering: A set of prerequisite toy tasks,” arXiv preprint
arXiv:1502.05698, 2015.
[34] M. Turan, Y. Almalioglu, E. P. Ornek, H. Araujo, M. F.
Yanik,
and M. Sitti, “Magnetic-visual sensor fusion-based dense 3d

reconstruction and localization for endoscopic capsule robots,”
arXiv preprint arXiv:1803.01048, 2018.
[35] M. Turan, Y. Y. Pilavci, R. Jamiruddin, H. Araujo, E.
Konukoglu,
and M. Sitti, “A fully dense and globally consistent 3d map
recon-
struction approach for gi tract to enhance therapeutic relevance
of the endoscopic capsule robot,” arXiv preprint
arXiv:1705.06524,
2017.
Sitti,
“Deep endovo: A recurrent convolutional neural network (rcnn)
based visual odometry approach for endoscopic capsule robots,”
Neurocomputing, vol. 275, pp. 1861–1870, 2018.
[37] M. Turan, E. P. Ornek, N. Ibrahimli, C. Giracoglu, Y.
Almali-
oglu, M. F. Yanik, and M. Sitti, “Unsupervised odometry and
depth learning for endoscopic capsule robots,” arXiv preprint
arXiv:1803.01047, 2018.
[38] W. Penfield, Mystery of the mind: A critical study of
consciousness
and the human brain. Princeton University Press, 2015.
[39] J. E. Swain, P. Kim, J. Spicer, S. Ho, C. J. Dayton, A.
Elmadih, and
K. Abel, “Approaching the biology of human parental attach-
ment: Brain imaging, oxytocin and coordinated assessments of
mothers and fathers,” Brain research, vol. 1580, pp. 78–101,
2014.
[40] C. Renzi, F. Leali, M. Cavazzuti, and A. Andrisano, “A

review
on artificial intelligence applications to the optimal design of
dedicated and reconfigurable manufacturing systems,” The In-
ternational Journal of Advanced Manufacturing Technology,
vol. 72,
no. 1-4, pp. 403–418, 2014.
[41] N. J. Nilsson, Principles of artificial intelligence. Morgan
Kauf-
mann, 2014.
[42] P. Gupta and P. Sahai, “A review on artificial intelligence
ap-
proach on prediction of software defects,” 2016.
[43] Y. LeCun, Y. Bengio, and G. Hinton, “Deep learning,”
nature, vol.
521, no. 7553, p. 436, 2015.
[44] S. Levine, C. Finn, T. Darrell, and P. Abbeel, “End-to-end
training
of deep visuomotor policies,” The Journal of Machine Learning
Research, vol. 17, no. 1, pp. 1334–1373, 2016.
[45] F. J. Varela, E. Thompson, and E. Rosch, The embodied
mind:
Cognitive science and human experience. MIT press, 2017.
[46] S. J. Russell and P. Norvig, Artificial intelligence: a
modern approach.
Malaysia; Pearson Education Limited,, 2016.
[47] Q. Kang, B. Huang, and M. Zhou, “Dynamic behavior of
artificial
hodgkin–huxley neuron model subject to additive noise,” IEEE
transactions on cybernetics, vol. 46, no. 9, pp. 2083–2093,

2016.
[48] V. Campos, A. Salvador, X. Giro-i Nieto, and B. Jou,
“Diving deep
into sentiment: Understanding fine-tuned cnns for visual senti-
ment prediction,” in Proceedings of the 1st International
Workshop
on Affect & Sentiment in Multimedia. ACM, 2015, pp. 57–62.
[49] S. Russell, D. Dewey, and M. Tegmark, “Research
priorities for
robust and beneficial artificial intelligence,” Ai Magazine, vol.
36,
no. 4, pp. 105–114, 2015.
[50] M. Turan, Y. Y. Pilavci, I. Ganiyusufoglu, H. Araujo,
E. Konukoglu, and M. Sitti, “Sparse-then-dense alignment-
based
3d map reconstruction method for endoscopic capsule robots,”
Machine Vision and Applications, vol. 29, no. 2, pp. 345–359,
2018.
[51] Z. Ghahramani, “Probabilistic machine learning and
artificial
intelligence,” Nature, vol. 521, no. 7553, p. 452, 2015.
[52] O. K. Oyedotun and A. Khashman, “Deep learning in
vision-
based static hand gesture recognition,” Neural Computing and
Applications, vol. 28, no. 12, pp. 3941–3951, 2017.
[53] R. K. Moore, “From talking and listening robots to
intelligent
communicative machines.”
[54] J. Sanchez-Riera, K.-L. Hua, Y.-S. Hsiao, T. Lim, S. C.

Hidayati,
and W.-H. Cheng, “A comparative study of data fusion for rgb-d
based visual recognition,” Pattern Recognition Letters, vol. 73,
pp.
1–6, 2016.
[55] J. Tang, C. Deng, and G.-B. Huang, “Extreme learning
machine
for multilayer perceptron,” IEEE transactions on neural
networks
and learning systems, vol. 27, no. 4, pp. 809–821, 2016.
[56] M. Abadi, A. Agarwal, P. Barham, E. Brevdo, Z. Chen, C.
Citro,
G. S. Corrado, A. Davis, J. Dean, M. Devin et al., “Tensorflow:
Large-scale machine learning on heterogeneous distributed sys-
tems,” arXiv preprint arXiv:1603.04467, 2016.
[57] G. E. Hinton, S. Osindero, and Y.-W. Teh, “A fast learning
algorithm for deep belief nets,” Neural computation, vol. 18,
no. 7,
pp. 1527–1554, 2006.
[58] C. Szegedy, V. Vanhoucke, S. Ioffe, J. Shlens, and Z.
Wojna,
“Rethinking the inception architecture for computer vision,” in
Proceedings of the IEEE Conference on Computer Vision and
Pattern
Recognition, 2016, pp. 2818–2826.
[59] L.-C. Chen, G. Papandreou, I. Kokkinos, K. Murphy, and
A. L.
Yuille, “Deeplab: Semantic image segmentation with deep con-
volutional nets, atrous convolution, and fully connected crfs,”

[60] J.-C. Chen, V. M. Patel, and R. Chellappa, “Unconstrained
face
verification using deep cnn features,” in Applications of
Computer
AUGUST 2015 11
Vision (WACV), 2016 IEEE Winter Conference on. IEEE, 2016,
pp.
1–9.
[61] C. Dong, C. C. Loy, K. He, and X. Tang, “Image super-
resolution
using deep convolutional networks,” IEEE transactions on
pattern
analysis and machine intelligence, vol. 38, no. 2, pp. 295–307,
2016.
[62] A. Gongal, S. Amatya, M. Karkee, Q. Zhang, and K. Lewis,
“Sensors and systems for fruit detection and localization: A
review,” Computers and Electronics in Agriculture, vol. 116,
pp. 8–
19, 2015.
[63] A. M. Nguyen, J. Yosinski, and J. Clune, “Innovation
engines:
Automated creativity and improved stochastic optimization via
deep learning,” in Proceedings of the 2015 Annual Conference
on
Genetic and Evolutionary Computation. ACM, 2015, pp. 959–
966.
[64] Y. Du, W. Wang, and L. Wang, “Hierarchical recurrent

neural
network for skeleton based action recognition,” in Proceedings
of
the IEEE conference on computer vision and pattern
recognition, 2015,
pp. 1110–1118.
[65] Y. Tang, “Deep learning using linear support vector
machines,”
[66] I. Stoica, D. Song, R. A. Popa, D. Patterson, M. W.
Mahoney,
R. Katz, A. D. Joseph, M. Jordan, J. M. Hellerstein, J. E.
Gonzalez
et al., “A berkeley view of systems challenges for ai,” arXiv
preprint arXiv:1712.05855, 2017.
[67] N. Sünderhauf, S. Shirazi, F. Dayoub, B. Upcroft, and M.
Milford,
“On the performance of convnet features for place recognition,”
in Intelligent Robots and Systems (IROS), 2015 IEEE/RSJ
Interna-
tional Conference on. IEEE, 2015, pp. 4297–4304.
[68] C. Chen, A. Seff, A. Kornhauser, and J. Xiao,
“Deepdriving:
Learning affordance for direct perception in autonomous driv-
ing,” in Computer Vision (ICCV), 2015 IEEE International
Conference
on. IEEE, 2015, pp. 2722–2730.
Sitti,
“A non-rigid map fusion-based rgb-depth slam method for en-
doscopic capsule robots,” arXiv preprint arXiv:1705.05444,

2017.
[70] N. Papernot, P. McDaniel, S. Jha, M. Fredrikson, Z. B.
Celik,
and A. Swami, “The limitations of deep learning in adversarial
settings,” in Security and Privacy (EuroS&P), 2016 IEEE
European
Symposium on. IEEE, 2016, pp. 372–387.
[71] M. Turan, Y. Almalioglu, H. Araujo, T. Cemgil, and M.
Sitti,
“Endosensorfusion: Particle filtering-based multi-sensory data
fusion with switching state-space model for endoscopic capsule
robots using recurrent neural network kinematics,” arXiv
preprint
arXiv:1709.03401, 2017.
[72] G. Huang, Y. Sun, Z. Liu, D. Sedra, and K. Q. Weinberger,
“Deep networks with stochastic depth,” in European Conference
on Computer Vision. Springer, 2016, pp. 646–661.
[73] T. P. Lillicrap, J. J. Hunt, A. Pritzel, N. Heess, T. Erez, Y.
Tassa,
D. Silver, and D. Wierstra, “Continuous control with deep rein-
forcement learning,” arXiv preprint arXiv:1509.02971, 2015.
[74] B. M. Lake, R. Salakhutdinov, and J. B. Tenenbaum,
“Human-
level concept learning through probabilistic program induction,”
Science, vol. 350, no. 6266, pp. 1332–1338, 2015.
[75] S. Marsland, “Machine learning, an algorithmic
perspective,
chapman & hall/crc machine learning & pattern recognition,”
CRC, Boca Raton, Fla, 2009.

[76] I. Lenz, H. Lee, and A. Saxena, “Deep learning for
detecting
robotic grasps,” The International Journal of Robotics Research,
vol. 34, no. 4-5, pp. 705–724, 2015.
[77] V. C. Müller and N. Bostrom, “Future progress in artificial
in-
telligence: A survey of expert opinion,” in Fundamental issues
of
artificial intelligence. Springer, 2016, pp. 555–572.
[78] J. S. Esteves, A. Carvalho, and C. Couto, “Generalized
geo-
metric triangulation algorithm for mobile robot absolute self-
localization,” in Industrial Electronics, 2003. ISIE’03. 2003
IEEE
International Symposium on, vol. 1. IEEE, 2003, pp. 346–351.
[79] A. Vedaldi and K. Lenc, “Matconvnet: Convolutional
neural
networks for matlab,” in Proceedings of the 23rd ACM
international
conference on Multimedia. ACM, 2015, pp. 689–692.
[80] A. Eitel, J. T. Springenberg, L. Spinello, M. Riedmiller,
and W. Bur-
gard, “Multimodal deep learning for robust rgb-d object recog-
nition,” in Intelligent Robots and Systems (IROS), 2015
IEEE/RSJ
International Conference on. IEEE, 2015, pp. 681–687.
[81] Y. Yang and Y. Li, “Robot learning manipulation action
plans
by” watching” unconstrained videos from the world wide web.”
2015.

[82] D. C. Cireşan, U. Meier, L. M. Gambardella, and J.
Schmidhuber,
“Deep, big, simple neural nets for handwritten digit
recognition,”
Neural computation, vol. 22, no. 12, pp. 3207–3220, 2010.
[83] J. Lu, V. Behbood, P. Hao, H. Zuo, S. Xue, and G. Zhang,
“Transfer
learning using computational intelligence: a survey,”
Knowledge-
Based Systems, vol. 80, pp. 14–23, 2015.
[84] J. Schmidhuber, “Deep learning in neural networks: An
overview,” Neural networks, vol. 61, pp. 85–117, 2015.
[85] H. Brighton and H. Selina, Introducing Artificial
Intelligence:
A Graphic Guide, ser. Introducing... Icon Books Limited,
2015. [Online]. Available: https://books.google.com.pk/books?
id=4GxGCgAAQBAJ
[86] J. Bai, Y. Wu, J. Zhang, and F. Chen, “Subset based deep
learning
for rgb-d object recognition,” Neurocomputing, vol. 165, pp.
280–
292, 2015.
[87] V. Veeriah, N. Zhuang, and G.-J. Qi, “Differential
recurrent neural
networks for action recognition,” in Computer Vision (ICCV),
2015
IEEE International Conference on. IEEE, 2015, pp. 4041–4049.
[88] M. F. Rooney and S. E. Smith, “Artificial intelligence in
engineer-
ing design,” Computers & Structures, vol. 16, no. 1-4, pp. 279–

288,
1983.
[89] Y. Duan, X. Chen, R. Houthooft, J. Schulman, and P.
Abbeel,
“Benchmarking deep reinforcement learning for continuous con-
trol,” in International Conference on Machine Learning, 2016,
pp.
1329–1338.
[90] S. Levine, P. Pastor, A. Krizhevsky, and D. Quillen,
“Learning
hand-eye coordination for robotic grasping with large-scale data
collection,” in International Symposium on Experimental
Robotics.
Springer, 2016, pp. 173–184.
[91] Y. Yang, C. Fermuller, Y. Li, and Y. Aloimonos, “Grasp
type
revisited: A modern perspective on a classical feature for
vision,”
in Proceedings of the IEEE Conference on Computer Vision and
Pattern
Recognition, 2015, pp. 400–408.
[92] C. S. Tirgul and M. R. Naik, “Artificial intelligence and
robotics,”
International Journal of Advanced Research in Computer
Engineering
& Technology, vol. 5, no. 6, pp. 1787–1793, 2016.
[93] S. E. Dilsizian and E. L. Siegel, “Artificial intelligence in
medicine
and cardiac imaging: harnessing big data and advanced comput-
ing to provide personalized medical diagnosis and treatment,”
Current cardiology reports, vol. 16, no. 1, p. 441, 2014.

[94] M. Imran, C. Castillo, J. Lucas, P. Meier, and S. Vieweg,
“Aidr:
Artificial intelligence for disaster response,” in Proceedings of
the
23rd International Conference on World Wide Web. ACM,
2014, pp.
159–162.
[95] M. Vrigkas, C. Nikou, and I. A. Kakadiaris, “A review of
human
activity recognition methods,” Frontiers in Robotics and AI,
vol. 2,
p. 28, 2015.
[96] H.-J. Yoo, “Deep convolution neural networks in computer
vi-
sion,” IEIE Transactions on Smart Processing & Computing,
vol. 4,
no. 1, pp. 35–43, 2015.
[97] N. Das, E. Ohn-Bar, and M. M. Trivedi, “On performance
evalua-
tion of driver hand detection algorithms: Challenges, dataset,
and
metrics,” in Intelligent Transportation Systems (ITSC), 2015
IEEE
18th International Conference on. IEEE, 2015, pp. 2953–2958.
[98] R. Salakhutdinov, “Learning deep generative models,”
Annual
Review of Statistics and Its Application, vol. 2, pp. 361–385,
2015.
[99] J. Schmidhuber, “On learning to think: Algorithmic
informa-

tion theory for novel combinations of reinforcement learning
controllers and recurrent neural world models,” arXiv preprint
arXiv:1511.09249, 2015.
[100] T. Chen, Z. Chen, Q. Shi, and X. Huang, “Road marking
detection
and classification using machine learning algorithms,” in
Intelli-
gent Vehicles Symposium (IV), 2015 IEEE. IEEE, 2015, pp.
617–621.
https://books.google.com.pk/books?id=4GxGCgAAQBAJ
https://books.google.com.pk/books?id=4GxGCgAAQBAJ1
Introduction2 Which of these and in what level can todayâ•Žs
artificial intelligence do?3 Where does the human intelligence
differ from AI?4 Why canâ•Žt we tell that todayâ•Žs AI is as
clever as human beings? 5 Give a comprehensive survey of what
can todayâ•Žs AI subfilelds do?6 Briefly explain technical
background as well?7 What is missing to todayâ•Žs AI still to
be called human level intelligence?8 Importance of Artificial
Intelligence9 Knowledge based towards understanding the
natural language10 Spatial frames of reference computation for
narrative text understanding11 What are open challenges?12
What are solution suggestions?13 What are future predictions
for AI?14 Which levels will it reach and which issues will be
solved in next decades?15 ConclusionReferences
MIRI
M A C H I N E I N T E L L I G E N C E
R E S E A R C H I N S T I T U T E
Advantages of Artificial Intelligences,
Uploads, and Digital Minds

Kaj Sotala
University of Helsinki, MIRI Research Associate
Abstract
I survey four categories of factors that might give a digital
mind, such as an upload or an
artificial general intelligence, an advantage over humans.
Hardware advantages include
greater serial speeds and greater parallel speeds. Self-
improvement advantages include
improvement of algorithms, design of new mental modules, and
modification of mo-
tivational system. Co-operative advantages include copyability,
perfect co-operation,
improved communication, and transfer of skills. Human
handicaps include compu-
tational limitations and faulty heuristics, human-centric biases,
and socially motivated
cognition. The shape of hardware growth curves, as well as the
ease of modifying minds,
are found to have a major impact on how quickly a digital mind
may take advantage of
these factors.
Sotala, Kaj. 2012. “Advantages of Artificial Intelligences,
Uploads, and Digital Minds.”
International Journal of Machine Consciousness 4 (1): 275–291.
doi:10.1142/S1793843012400161.
This version contains minor changes.
http://dx.doi.org/10.1142/S1793843012400161

Kaj Sotala
1. Introduction
A digital mind is a mind that runs on a computer. One type of
digital mind is a mind
upload, a hypothetical mind that was originally human, but that
has been moved into a
digital format and is being run as a software program on a
computer. Another type of
digital mind is that of an artificial general intelligence (AGI).
While uploads are based
on taking existing human minds and closely replicating them in
software, AGI may be
built on computer science principles and have little or no
resemblance to the human
psyche.
Either type of a digital mind might be created within a
timeframe of decades to cen-
turies. A recent roadmap charting the technological
requirements for creating uploads
suggests that they may be feasible by mid-century (Sandberg
and Bostrom 2008). So-
tala and Valpola (2012) note that research into prostheses
replicating the functions of the
hippocampus and the cerebellum are well under way, and
suggest that a feasible future
development would be “exocortices,” implants that can be
connected to human brains
and which gradually take over cortical brain function.
Interest in AGI research is also growing, with an increasing
number of special ses-
sions, workshops, and conferences devoted specifically to topics
such as AGI having been

held in the recent years (Baum, Goertzel, and Goertzel 2011). In
an expert assessment
survey conducted at the Artificial General Intelligence 2009
(AGI-09) conference, the
median estimates for when there would be a 10%, 50%, or a
90% chance of having an
AGI capable of passing the third grade were 2020, 2030 and
2075, respectively (Baum,
Goertzel, and Goertzel 2011). In an informal survey conducted
at the 2011 Winter
Intelligence Conference, the median estimates for when there
would be a 10%, 50%,
or a 90% chance of developing human-level machine
intelligence were 2028, 2050, and
2150 (Sandberg and Bostrom 2011).
There has been some previous work focused on examining the
consequences of cre-
ating digital intelligences. Focusing specifically on uploads,
researchers have examined
some of the economic consequences of the ability to copy minds
(Hanson 1994, 2008),
as well as the improved coordination ability stemming from
being able to copy, delete
and restore minds (Shulman 2010). Hanson (1998) and Kaas et
al. (2010) look more
generally at the economic effects of digital minds that can be
copied. Sotala and Valpola
(2012) explore the way that mind uploading may lead to “mind
coalescence,” the ability
to merge previously separate minds together.
Other researchers have argued that once we have AGI, it will
surpass the capabilities
of humans in many domains at drastic speed (see e.g. Vinge
[1993], Bostrom [2003],

Yudkowsky [2008a], and Chalmers [2010]). A “hard takeoff ”
(Yudkowsky 2001; Bugaj
and Goertzel 2007; Hall 2008; Vinge 2008) involves an AGI
reaching a point which
1
Minds
allows it to quickly accumulate various advantages and
influence, becoming a dominant
power before humans have the time to properly react. Should
existing human prepara-
tion be inadequate for such a drastic event, there could be
serious consequences, up to
and including human extinction (Bostrom 2002; Yudkowsky
2008a; Chalmers 2010).
This paper attempts to study the consequences of creating a
digital mind in terms of the
advantages that they may enjoy over humans.
1.1. Intelligence, Optimization Power, and Advantages
A great variety of definitions have been offered for the word
“intelligence.” A definition
which seems to summarize the essential content in most of them
is that intelligence
measures an agent’s ability to achieve goals in a wide range of
environments (Legg and
Hutter 2007). A mind has goals which it tries to achieve, and
more intelligent minds
are better at finding, inventing, and evaluating various ways of
achieving their goals. A

generalization of the concept of intelligence is the notion of
optimization power (Yud-
kowsky 2008a; Muehlhauser and Helm 2012), an agent’s general
ability to achieve its
goals. While intelligence is derived from what are generally
considered “mental” fac-
ulties, an agent’s optimization power is also a factor of things
such as its allies and re-
sources, as well as its ability to obtain more of them.
The crucial risk involved in the creation of digital minds is the
possibility of creating
minds whose goals are very different from humanity’s, and who
end up possessing more
optimization power than humanity does. Current human
preferences and desires seem
to be very complex and not well-understood: there is a strong
possibility that only a very
narrow subset of all possible goals will, if successful, lead to
consequences that would be
considered favorable by humans (Yudkowsky 2008a;
Muehlhauser and Helm 2012). If
digital intelligences are created and end up having (as a group)
more optimization power
than humanity does, and their goals are very different from
humanity’s goals, then the
consequences are likely to be considered very bad by most
humans.
This paper attempts to analyze the consequences of creating a
digital mind from the
perspective of the optimization power that they might
accumulate. Factors that may
lead to digital minds accumulating more optimization power
than humans are called
advantages; factors that may lead to digital minds accumulating

less optimization power
than humans are called disadvantages.
2. Hardware Advantages
A digital mind running on a computer system can upgrade the
system to utilize more
powerful hardware, while biological humans cannot drastically
upgrade their brains.
Suppose that there is some minimum hardware configuration
that provides a digital
2
Kaj Sotala
mind with roughly the same processing power and memory as a
human brain. Any
increase in hardware resources past this point is a hardware
advantage in favor of the
digital mind.
2.1. Superior Processing Power
The amount of processing power required to run a digital mind
is currently unknown.1
For uploads, Sandberg and Bostrom (2008) place 1018 to 1025
FLOPS as the most likely
amount required to run one in real time. They estimate that
current trends would make
these levels available for purchase at the cost of 1 million
dollars around 2019 to 2044.

An upload attempts to accurately emulate the entirety of human
brain function, with
all relevant details intact. In contrast, AGI designers are free to
use any working algo-
rithm, regardless of its biological plausibility. The human brain
is evolved to function in
a way suited to the constraints of biology, which may be very
different from what would
run efficiently on a computer. This allows for the possibility
that an AGI would require
much less processing power than an upload. Estimates of the
amount of processing
power required to run a mind range from modern-day computers
(Hall 2007), to 1011
FLOPS (Moravec 1998), and 1014 FLOPS (Bostrom 1997).
2.1.1. Superior Serial Power
Humans perceive the world on a particular characteristic time
scale. A mind being
executed on a system with greatly superior serial power could
run on a faster timescale
than we do. For instance, a mind with twice the serial power of
the human brain might
experience the equivalent of two seconds passing for each
second that we did, thinking
twice the amount of thoughts in the same time. This advantage
would be especially
noticeable in time-critical decision-making.
Even a small advantage would accumulate given enough time.
Over the course of a
year, a 10% difference in speed would give the faster mind more
than an extra month.
This would allow it to outcompete any mind with equal skills

and resources but without
the speed advantage.
In the “speed explosion” scenario (Solomonoff 1985;
Yudkowsky 1996; Chalmers
2010), digital researchers, running at an accelerated speed, work
to develop faster com-
puters. If the minds doing the research could take advantage of
the faster hardware
they produced, the time required to develop the next generation
of hardware could keep
1. Some sources provide their estimates in terms of MIPS
(Millions of Instructions Per Second), while
others use FLOPS (Floating-Point Operations per Second).
These are not directly comparable, and there
is no reliable way to convert between the two. For this paper,
we have used the rough estimate in Sandberg
and Bostrom (2008) that FLOPS grow as MIPS to the power of
0.8. The authors warn that this trend
may change, with the exponent possibly becoming larger than 1
3
Minds
getting shorter as the researchers would be getting more done in
the same time. This
could continue until some bottleneck, such as the time needed to
physically build the
computers, or fundamental barrier was reached.
2.1.2. Increased Parallel Power, Increased Memory

Recent advances in computing power have been increasingly
parallel instead of serial.
If the trend keeps up, a future computer may not be able to use
its superior processing
power to gain a direct increase in speed over the human brain, if
the tasks in question do
not parallelize well. Amdahl’s law states that if a fraction f of a
program’s performance
can be parallelized, then the speedup given by n processors
instead of one is 1/((1−f)+
f/n) (Amdahl 1967). A difference of several orders of magnitude
in computing power
might translate to a much more modest change in speed.
Gustafson (1988) notes that
in practice, it is the parallelizable part of a problem that grows
as data is added, and the
serial part remains constant. Even if increasing the number of
processors didn’t allow
a problem to be solved in less time, it can allow a larger
problem to be solved in better
detail.
As the human brain works in a massively parallel fashion, at
least some highly parallel
algorithms must be involved with general intelligence. Extra
parallel power might then
not allow for a direct improvement in speed, but it could
provide something like a greater
working memory equivalent. More trains of thought could be
pursued at once, and
more things could be taken into account when considering a
decision. Brain size seems
to correlate with intelligence within rats (Anderson 1993),
humans (McDaniel 2005),
and across species (Deaner et al. 2007), suggesting that

increased parallel power could
make a mind generally more intelligent.
3. Self-Improvement Advantages
A digital mind with access to its source code may directly
modify the way it thinks,
or create a modified version of itself. In order to do so, the
mind must understand its
own architecture well enough to know what modifications are
sensible. An AGI can
intentionally be built in a manner that is easy to understand and
modify, and may even
read its own design documents. Things may be harder for
uploads, especially if the
human brain is not yet fully understood by the time uploading
becomes possible.
Either type of mind could experiment with a large number of
possible interventions,
creating thousands or even millions of copies of itself to see
what kinds of effects various
modifications have. While some of the modifications could
produce unseen long-term
problems, each copy could be subjected to various intensive
tests over an extended period
of time to estimate the effects of the modifications. Copies with
harmful or neutral
4
Kaj Sotala
modifications could be deleted, making room for alternative

ones. (Shulman 2010).
Less experimental approaches might involve formal proofs of
the effects of the changes
to be made.
Recursive self-improvement (Yudkowsky 2008a; Chalmers
2010) is a situation in
which a mind modifies itself, which then makes it capable of
further improving itself.
For instance, an AGI might improve its pattern-recognition
capabilities, which would
then allow it to notice inefficiencies in itself. Correcting these
inefficiencies would free
up processing time and allow the AGI to notice more things that
could be improved.
To a limited extent, humans have been engaging in recursive
self-improvement as
the development of new technologies and forms of social
organization have made it
possible to organize better and develop yet more advanced
technologies. Yet the core
of the human brain has remained the same. If changes could be
found that sparked
off more changes, which kept sparking off more changes, the
result could be a greatly
improved form of intelligence (Yudkowsky 2008a).
3.1. Improving Algorithms
A digital mind could come across algorithms in itself that could
be be improved. For
instance, they could be made faster, to consume less memory, or
to rely on fewer assump-
tions. In the simplest case, an AGI implementing some standard
algorithm might come

across a paper detailing an improved implementation of it. Then
the old implementation
could be simply replaced with the new one. An upload with
emulated neurons might
alter itself so as to mimic the effects of drugs, neurosurgery,
genetic engineering and
other interventions (Shulman 2010).
In the past, improvements in algorithms have sometimes been
even more important
than improvements in hardware. The President’s Council of
Advisors on Science and
Technology (PCAST 2010) mentions that performance on a
benchmark production
planning model improved by a factor of 43 million between
1988 and 2003. Out of
the improvement, a factor of roughly 1,000 was due to better
hardware and a factor of
roughly 43,000 was due to improvements in algorithms. Also
mentioned is an algorith-
mic improvement of roughly 30,000 for mixed integer
programming between 1991 and
2008.
3.2. Designing New Mental Modules
A mental module, in the sense of functional specialization
(Cosmides and Tooby 1994;
Barrett and Kurzban 2006), is a part of a mind that specializes
in processing a certain
kind of information. Specialized modules are much more
effective than general-purpose
ones, for the number of possible solutions to a problem in the
general case is infinite.
Research in a variety of fields, including artificial intelligence,
developmental psychol-

5
Minds
ogy, linguistics, perception and semantics has shown that a
system must be predisposed
to processing information within the domain in the right way or
it will be lost in the sea
of possibilities (Tooby and Cosmides 1992). Many problems
within computer science
are intractable in the general case, but can be efficiently solved
by algorithms customized
for specific special cases with useful properties that are not
present in general (Cormen
et al. 2009). Correspondingly, many specialized modules have
been proposed for hu-
mans, including modules for cheater-detection, disgust, face
recognition, fear, intuitive
mechanics, jealousy, kin detection, language, number, spatial
orientation, and theory of
mind (Barrett and Kurzban 2006).
Specialization leads to efficiency: to the extent that regularities
appear in a problem,
an efficient solution to the problem will exploit those
regularities (Kurzban 2010). A
mind capable of modifying itself and designing new modules
customized for specific
tasks might eventually outperform biological minds in any
domain, even presuming no
hardware advantages. In particular, any improvements in a
module specialized for cre-

ating new modules would have a disproportionate effect.
It is important to understand what specialization means in this
context, for several
competing interpretations exist. For instance, Bolhuis et al.
(2011) argue against func-
tional specialization in nature by citing examples of “domain-
general learning rules” in
animals. On the other hand, Barrett and Kurzban (2006) argue
that even seemingly
domain-general rules, such as the modus ponens rule of formal
logic, operate in a re-
stricted domain: representations in the form of if-then
statements. This paper uses Bar-
rett and Kurzban’s broader interpretation. Thus, in defining the
domain of a module,
what matters is not the content of the domain, but the formal
properties of the pro-
cessed information and the computational operations performed
on the information.
Positing functional modules in humans also does not imply
genetic determination, nor
that the modules could necessarily be localized to a specific
part of the brain (Barrett
and Kurzban 2006).
A special case of a new mental module is the design of a new
sensory modality, such
as that of vision or hearing. Yudkowsky (2007) discusses the
notion of new modalities,
and considers the detection and identification of invariants to be
one of the defining
features of a modality. In vision, changes in lighting conditions
may entirely change the
wavelength of light that is reflected off a blue object, but it is
still perceived as blue. The

sensory modality of vision is then concerned with, among other
things, extracting the
invariant features that allow an object to be recognized as being
of a specific color even
under varying lighting.
Brooks (1987) mentions invisibility as an essential difficulty in
software engineering.
Software cannot be visualized in the same way physical
products can be, and any visual-
ization can only cover a small part of the software product.
Yudkowsky (2007) suggests
6
Kaj Sotala
a codic cortex designed to model code the same way that the
human visual cortex is
evolved to model the world around us. Whereas the designer of
a visual cortex might
ask “what features need to be extracted to perceive both an
object illuminated by yellow
light and an object illuminated by red light as ‘the color blue’?”
the designer of a codic
cortex might ask “what features need to be extracted to perceive
the recursive algorithm
for the Fibonacci sequence and the iterative algorithm for the
Fibonacci sequence as
‘the same piece of code’?” Speculatively, new sensory
modalities could be designed for
various domains for which existing human modalities are not
optimally suited.

3.3. Modifiable Motivation Systems
Humans frequently suffer from problems such as
procrastination, boredom, mental fa-
tigue, and burnout. A mind which did not become bored or tired
with its work would
have a clear advantage over humans. Shulman (2010) notes two
ways by which uploads
could overcome these problems. Uploads could be copied while
they were in a rested
and motivated state. When they began to tire, they could be
deleted and replaced with
the “snapshot” taken while they were still rested. Alternatively,
their brain state could
be edited so as to eliminate the neural effects of boredom and
fatigue. An AGI might
not need to have boredom built into it in the first place.
The ability for a mind to modify its own motivational systems
also has its own risks.
Wireheading (Yudkowsky 2001; Omohundro 2008) is a
phenomenon where a mind
self-modifies to make it seem like it is achieving its goals, even
though it is not. For
instance, an upload might try to eliminate its stress about its
friends dying by creating a
delusion about them always being alive. Once a mind has
wireheaded, it may no longer
want to fix its broken state.
Even if wireheading-related problems were avoided, a mind
altering its own moti-
vations still risks an outcome where its ability to pursue its
original goals is worsened.
To avoid such problems, a mind might attempt to formally
prove that proposed changes

do not alter its current goals (Yudkowsky 2008a), or it may
produce modified copies of
itself and subject the copies to an intensive testing regimen
(Shulman 2010).
4. Co-operative Advantages
4.1. Copyability
A human child takes nine months to gestate, after which close
to two or even three
decades are typically needed before it can do productive work,
depending on the type
of work. Raising a child is expensive, costing on average
between 216,000 and 252,000
dollars over 18 years in the United States (Lino 2010). In
contrast, a digital mind can be
copied very quickly and doing so has no cost other than access
to the hardware required
7
Minds
to run it. Hanson (1994, 2008) estimates that copyable workers
could rapidly come
to dominate major portions of the economy, as the mind willing
to work for the lowest
wage could be copied until there was no more demand for
workers of that type. Although
such workers would be individually poor, they would control a
large amount of wealth
as a group. Whether they could take advantage of it would

depend on their ability to
co-operate.
For human populations, the maximum size of the population
depends primarily on
the availability of resources such as food and medicine. For
populations of digital minds,
the maximum size of the population depends primarily on the
amount of hardware avail-
able. A digital mind could obtain more hardware resources by
legitimately buying them,
or by employing illegal means such as hacking and malware.
Modern-day botnets are networks of computers that have been
compromised by out-
side attackers and are used for illegitimate purposes. Estimates
of their size range from
one study saying the effective sizes of botnets rarely exceed a
few thousand bots, to a
study saying that botnet sizes can reach 350,000 members
(Rajab et al. 2007). Cur-
rently, the distributed computing project [email protected], with
290,000 active clients,
can reach speeds in the 1015 FLOPS range ([email protected]
2012). A relatively conser-
vative estimate that presumed a digital mind couldn’t hack into
more computers than
the best malware practitioners of today, that a personal
computer would have a hundred
times the computing power of today, and that the mind took
1013 FLOPS to run, would
suggest that a single mind could spawn 12,000 copies of itself.
Resorting to illegal means may not be necessary if digital minds
are allowed to own
property, or at least earn money. Uploads might be legally

considered the same person
as before the upload. AGIs owned by a company or private
individual can accumulate
property for their owner, or they may try to set up a front
company to act through.
Creating new copies of a mind is profitable until the costs of
maintaining a new copy
exceed the profits that copy is capable of generating, so copies
might be created until the
wage a copy can earn falls to the level of maintaining the
hardware (Hanson 1994, 2008).
Hanson (2008) argues that copying would drive wages down to
machine-subsistence
levels, leading to “insectlike urban densities, with many billions
[of digital minds] living
in the volume of a current skyscraper, paying astronomical rents
that would exclude most
humans.”
For uploads, who might not be capable of co-operating with
each other any better
than current-day humans do, this might be a disadvantage rather
than an advantage.
But if they could, or if an AGI spawned many copies of itself,
then the group could pool
their resources and together control a large fraction of the
wealth in the world.
8
Kaj Sotala
4.2. Perfect Co-operation

Human capability for co-operation is limited by the fact that
humans have their own
interests in addition to those of the group. Olson (1965) showed
that it is difficult for
large groups of rational, selfish agents to effectively co-operate
even to achieve common
goals, if those goals can be characterized as obtaining a public
good for the group. Public
goods are those that, once obtained, benefit everyone and
cannot effectively be denied to
anyone. Because an individual’s contribution has a negligible
effect on the achievement
of the good, all group members have an incentive to free ride on
the effort.
One implication is that smaller groups often have an advantage
over larger ones.
For one, co-operation is easier to enforce in a smaller group.
Additionally, it may be
beneficial for, e.g., a large company to lobby for laws
benefiting the whole industry,
since it is large enough to benefit from the laws even if it had to
shoulder almost all of
the costs of lobbying itself. Smaller companies forgo such
lobbying and free ride on the
large company’s investment. This leads to lobbying investment
that is suboptimal for the
industry as a whole (Olson 1965; Mueller 2003). Self-interest is
a natural consequence
of evolution, as it increases the odds that an organism survives
long enough to breed.
Drives such as self-preservation are also natural instrumental
values for any intelligent
agent, for only entities that continue to exist can work towards
their goals (Omohundro
2008).

However, minds might be constructed to lack any self-interest,
particularly if mul-
tiple copies of the same mind existed and the destruction of one
would not seriously
threaten the overall purpose. Such entities’ minds could be
identical to one another,
share the same goal system, and co-operate perfectly with one
another with no costs
from defection or from systems for enforcing co-operation.
In the case of uploads, this could happen through copying an
upload in suitable ways,
creating a “superorganism.” An upload that has been copied
may hold the view that be-
ing deleted is an acceptable cost to pay, for as long as other
copies of it survive. Uploads
holding this view would then be ready to make large sacrifices
for the rest of the su-
perorganism, and might employ various psychological
techniques to reinforce this bond
(Shulman 2010). Minds wishing to work for a common goal
might also choose to con-
nect their brains together, more or less coalescing together into
a single mind (Sotala and
Valpola 2012). A lighter form of mind coalescence might also
be used to strengthen the
unity of a superorganism.
4.3. Superior Communication
Misunderstandings are notoriously common among humans.
AGI could potentially
need to spend much less effort on communication. Language can
be thought of as sym-
bols that map to different individual’s conceptual spaces, with

miscommunication oc-
9
Minds
curring because of different mappings (Honkela et al. 2008).
Efficiency of communica-
tion could be improved by having very similar conceptual
spaces (aiding communication
between copies) or via custom-tailoring mental modules to the
domain of conceptual
mapping. Such modules could e.g. simulate the interpretations
to different messages
emerging from a wide variety of conceptual spaces and seek to
include the caveats ex-
cluding those interpretations, or directly communicate parts of
the presumed conceptual
spaces using some standard language.
Humans are limited as to the speed at which they can listen to
others, or read written
text, without loss of comprehension. Increased skill at doing
conceptual mapping, as
well as increased processing power, could plausibly increase
these rates. Digital minds
could also communicate at higher bandwidths, transmitting a
vastly larger amount of
information at once. If necessary, minds could join their minds
together, exchanging
thoughts directly (Sotala and Valpola 2012).
4.4. Transfer of Skills

Copying parts of a mind is a special case of copying a whole
mind. To the extent that
skills can be modularized, digital minds could create self-
contained skill modules to be
shared with others. In the most extreme case, a population of
minds could outsource all
of their skills to a very small number of cognitive modules, only
learning a small number
of things themselves (Bostrom 2004). Whenever one mind had
mastered a skill, it could
share it with all the others.
5. Human Handicaps
Humans frequently reason in suboptimal or incorrect ways
(Stanovich 2008; Kahneman
2011). Such failures of reasoning have an enormous negative
impact on society. Among
other things, they cause people to suffer from a worse standard
of living, make bad
investments, become more easily manipulated, end up falsely
accused or imprisoned by
the authorities, increase the mortality rate, or even fall prey to
scams serious enough to
crash a national economy (Stanovich 2008). A mind that was
immune to such biases
would reason more reliably than we do, while possibly
exploiting our biases. An upload
may attempt to self-modify to overcome its biases, while an
AGI might never have the
biases in the first place.
5.1. Biases from Computational Limitations or False
Assumptions

Some human biases can be seen as assumptions or heuristics
that fail to reason correctly
in a modern environment, or as satisfying algorithms that do the
best possible job given
human computational resources (Gigerenzer and Brighton
2009). A digital mind could
10
Kaj Sotala
potentially overcome most if not all of the biases that plague
human reasoning, either
by rewriting its algorithms to better suit the environment or to
better take into account
growing computational resources.
For example, when faced with a difficult question, human brains
have a tendency to
instead solve an easier question and treat it as the answer to the
more difficult question.
If asked “how much will this company have grown in five
years,” an intuitive answer
might be generated based on the question “how fast has the
company been growing so
far.” These kinds of heuristics often function well, but they
sometimes fail to take into
account crucial factors, such as reasons for why the company
might fail to keep up its
historical growth rates. While such factors can be taken into
account by explicitly think-
ing about them, it requires explicit thought and is not done
automatically (Kahneman
2011). A digital mind might be capable of editing the heuristics

it uses for answering
such questions, and avoid the risk of accepting the heuristic
answer as a fact without
further evaluation, like humans often do.
One’s susceptibility to some major biases, such as
overconfidence and hindsight bias,
correlates negatively with one’s general intelligence. This
suggests that computational
limitations cause at least some of the flaws in human reasoning
(Stanovich and West
1998).
5.2. Human-Centric Biases
People tend to think of the capabilities of non-human minds,
such as God or an artificial
intelligence, as if the minds in question were human. This
tendency persists even if
humans are explicitly instructed to act otherwise (Barrett and
Keil 1996). This is a
special case of biases due to false assumptions.
Evolutionarily, other minds have constituted possibly the single
most important se-
lection pressure facing any single human—the extent to which
one can co-operate with
others and avoid being exploited will to a large extent
determine one’s success in life. Be-
cause mental states such as beliefs, motives, intentions, and
emotions cannot be directly
observed and have to be inferred, we are likely to have evolved
a large number of algo-
rithms and modules for inferring such states on the basis of very
subtle cues (Cosmides
and Tooby 1994). Since we have never had to model the

thoughts of non-human minds
to this extent, these modules will automatically try to model any
minds we’re dealing
with using the same principles. Thus they will carry over
human-centric assumptions
when we attempt to model the behavior of non-human minds
(Yudkowsky 2008a).
To some extent, the modules may even assume we’re dealing
with humans similar
to ourselves: neural systems we use for modeling others overlap
with those used for
processing information about ourselves (Uddin et al. 2007).
Humans in a “cold” (non-
emotional, non-aroused) state frequently overestimate their
degree of self-control in a
11
Minds
“hot” (emotional, aroused) state (Loewenstein 2005). Even a
relatively mild difference
between oneself and the mind to be modeled can thus lead to
erroneous predictions.
To the extent that digital minds reason and behave unlike
humans, our attempts to
intuitively predict their behavior will be based on false
premises. To some extent, this
disadvantage may be symmetrical, in that e.g., uploads engaged
in self-modification will
fail to understand how non-modified humans would behave.

AGIs may not start out
with any efficient models for human behavior in the first place.
Over time, biological
humans will accumulate expertise in predicting digital minds
and digital minds may
learn and self-modify to better understand humans. However,
biological humans may
be at a disadvantage if self-modification is easy, for the
assumptions applying to digital
minds may change faster than humans can keep up with.
5.3. Biases from Socially Motivated Cognition
It has also been proposed that humans have evolved to acquire
beliefs which are socially
beneficial, even if those beliefs weren’t true (Trivers 2000;
Kurzban and Aktipis 2007;
Kurzban 2010). For instance, minds may be biased to believe
that they’ll be successful in
order to persuade others to ally with them. Even more seriously,
minds may not be built
to actually question the accuracy of their beliefs, but to
rationalize reasonable-sounding
explanations for their initial emotional reaction to a concept. If
human reasoning is
strongly enough biased to come up with popular or self-
beneficial theories, instead of
theories that are actually true, a mind without such a bias could
be immensely more
effective at reaching the correct theories.
6. Discussion
Two main questions seem to emerge:
What do hardware growth curves look like?

A number of advantages are either completely based on
improved hardware (superior
processing power) or made greatly stronger by it (overcoming
biases, designing new
mental modules, copyability, superior communication).
Therefore the faster hardware
advances, the steeper the takeoff.
Digital minds are subject to the risk of hardware overhang
(Yudkowsky 2008b; Shul-
man and Sandberg 2010). If software development proceeds
slower than hardware de-
velopment, then the hardware required for digital minds may be
available far earlier than
the software. When the software for digital minds is developed,
the minds could have
at their disposal much more hardware resources than is strictly
necessary to run them,
giving them an unexpected advantage.
12
Kaj Sotala
Even if hardware development stopped for a while and digital
minds were stuck on a
certain level that put them on roughly equal footing with
humanity, this situation could
not be relied upon to persist. Any future breakthrough in
hardware would have the
potential to upset the situation and give the digital minds a
decisive advantage. Lloyd
(2000) estimates the ultimate physical limits on hardware to

allow for a one-liter, one-
kilogram computer capable of carrying out 1050 OPS if we
allow for computers made
of exotic matter that explode during the calculation, or 1040
OPS if we constrain our-
selves to computers made of ordinary matter. Whatever estimate
we use for the human
brain’s processing power, we cannot with any certainty presume
that it would be near
the achievable physical limits on computation.
How modifiable are minds?
A number of advantages (improved parallel power, overcoming
biases, designing new
mental modules, perfect co-operation, superior communication,
transfer of skills) rely
to differing degrees on the assumption that digital minds are
easy to understand and
modify. To the degree to which this assumption is untrue, these
advantages become
less pronounced. Loosemore (2007) argues that there may be a
disconnect between the
local behavior of interacting elements and the global behavior
of the system in a mind,
so that generally intelligent behavior might not be derivable
from mathematical rules.
This would reduce the pace of self-improvement, though self-
improvement would still
be possible via systematic exploration of related mental
architectures.
A closely related and important question is the “intelligibility
of intelligence” (Sala-
mon, Rayhawk, and Kramár 2010)—the question of whether the
core of general intel-

ligence could be expressed in a compact intelligible theory, like
the theory of relativity,
or whether it is more akin to a “swiss army knife” of
incremental solutions and special
tricks. If intelligence is generally unintelligible and hard, then
improving upon minds
might prove difficult and slow.
Bach (2010) argues that like AGIs, human organizations such as
corporations, ad-
ministrative and governmental bodies, churches and universities
are intelligent agents
that are more powerful than individual humans, and that the
development of AGI would
increase the power of organizations in a quantitative way but
not cause a qualitative
change.
Humans grouping into organizations are to some degree capable
of taking advantage
of increased parallel (but not serial) speed by adding more
individuals. While organi-
zations can institute guidelines such as peer review that help
combat bias, working in
an organization can introduce biases of its own, such as
groupthink (Esser 1998). They
cannot design new mental modules or benefit from any of the
co-operative advantages
digital minds may enjoy. Possibly their largest shortcoming is
their reduced efficiency as
13

Minds
the size of the organization grows and their general
susceptibility to having their original
goals hijacked by smaller interest groups within the
organization (Olson 1965).
7. Conclusions
Digital minds potentially enjoy a number of advantages, all of
which might make it
easier for them to succeed in their goals. Hardware
improvements could allow digital
minds to think faster and to consider more things at once. Self-
improvement advan-
tages would allow digital minds to modify themselves, possibly
in a recursive fashion
where initial improvements kept sparking off further
improvements. Algorithms could
be improved to give an even larger boost than hardware
improvements could, new men-
tal modules could be designed for new kinds of domains, and
the various motivational
systems plaguing humans could be overcome. Co-operative
advantages include copy-
ability, which would let a mind replicate itself many times over,
and the potential for
perfect co-operation would eliminate conflict between copies of
the same mind. Supe-
rior communication and the ability to transfer skills would
further help matters. Finally,
humans may suffer from various handicaps in their mental
architecture. They might
employ inadequate heuristics, have a harder time modeling
digital minds than digital
minds have modeling humans, and suffer from socially

motivated cognition.
The extent to which these potential advantages could be realized
is an open question.
Hardware advantages rely on hardware progressing, copyability
relies on there being
enough hardware to run a large number of minds, and nearly all
of the others rely on
minds being sufficiently modifiable.
It is possible that hard takeoff scenarios have gotten a
disproportionate amount of
attention in discussions of digital mind advantages. From a
safety viewpoint, assuming
that a digital mind can bootstrap itself to superintelligence in a
matter of weeks or hours
is a conservative guess, in the sense that it’s the scenario that
leaves others the least
time to prepare (Yudkowsky 2001, 2008a). We should try to
anticipate such a scenario,
because if we do not, there will be little or no time to react
when it happens. Yet debates
over the plausibility of a hard takeoff distract from the fact that
digital minds developing
over a timespan of years or decades is a dangerous scenario as
well. Digital minds might
be developed, then be relatively weak for an extended time,
until some hardware or
software breakthrough suddenly allowed them to become
considerably more powerful.
Likewise, coalitions could initially form, keeping each other in
check, until something
happened to break the balance. The possibility of a hard takeoff
is definitely real and
deserves attention, but it is far from the only danger.

14
Kaj Sotala
Acknowledgments
This paper began as a collaboration with Zack M. Davis, who
then withdrew from the
project. Regardless, this paper might not have existed without
his early work.
The author would also like to thank Alexandros Marinos, Anna
Salamon, Louie
Helm, several pseudonymous commenters on lesswrong.com,
two anonymous review-
ers, and anyone else he may have forgotten but who did not
deserve to be forgotten, for
their feedback on the article.
References
Amdahl, Gene M. 1967. “Validity of the Single Processor
Approach to Achieving Large Scale Computing
Capabilities.” In Proceedings of the April 18–20, 1967, Spring
Joint Computer Conference—AFIPS ’67
(Spring), 483–485. New York: ACM Press.
doi:10.1145/1465482.1465560.
Anderson, Britt. 1993. “Evidence from the Rat for a General
Factor That Underlies Cognitive Perfor-
mance and That Relates to Brain Size: Intelligence?”
Neuroscience Letters 153 (1): 98–102. doi:10.
1016/0304-3940(93)90086-Z.

Bach, James. 2010. “How the Singularity of Artificial
Intelligence Might Be Achieved, and Why it Does
Not Matter.” In Mainzer 2010, 471–475.
Barrett, H. Clark, and Robert Kurzban. 2006. “Modularity in
Cognition: Framing the Debate.” Psycho-
logical Review 113 (3): 628–647. doi:10.1037/0033-
295X.113.3.628.
Barrett, Justin L., and Frank C. Keil. 1996. “Conceptualizing a
Nonnatural Entity: Anthropomorphism
in God Concepts.” Cognitive Psychology 31 (3): 219–247.
doi:10.1006/cogp.1996.0017.
Baum, Seth D., Ben Goertzel, and Ted G. Goertzel. 2011. “How
Long Until Human-Level AI? Results
from an Expert Assessment.” Technological Forecasting and
Social Change 78 (1): 185–195. doi:10.
1016/j.techfore.2010.09.006.
Bolhuis, Johan J., Gillian R. Brown, Robert C. Richardson, and
Kevin N. Laland. 2011. “Darwin in
Mind: New Opportunities for Evolutionary Psychology.” PLoS
Biology 9 (7): e1001109. doi:10 .
1371/journal.pbio.1001109.
Bostrom, Nick. 1997. “Predictions from Philosophy? How
Philosophers Could Make Themselves Useful.”
Unpublished manuscript. Last revised September 19, 1998. http
: / / www . nickbostrom . com /
old/predict.html.
. 2002. “Existential Risks: Analyzing Human Extinction
Scenarios and Related Hazards.” Journal
of Evolution and Technology 9.
http://www.jetpress.org/volume9/risks.html.

. 2003. “When Machines Outsmart Humans.” Futures 35 (7):
759–764. doi:10.1016/S0016-
3287(03)00026-0.
. 2004. “The Future of Human Evolution.” In Two Hundred
Years After Kant, Fifty Years After
Turing, edited by Charles Tandy, 2:339–371. Death and Anti-
Death. Palo Alto, CA: Ria University
Press.
Brooks, Frederick P., Jr. 1987. “No Silver Bullet: Essence and
Accidents of Software Engineering.” Com-
puter 20 (4): 10–19. doi:10.1109/MC.1987.1663532.
15
http://lesswrong.com
http://dx.doi.org/10.1145/1465482.1465560
http://dx.doi.org/10.1016/0304-3940(93)90086-Z
http://dx.doi.org/10.1016/0304-3940(93)90086-Z
http://dx.doi.org/10.1037/0033-295X.113.3.628
http://dx.doi.org/10.1006/cogp.1996.0017
http://dx.doi.org/10.1016/j.techfore.2010.09.006
http://dx.doi.org/10.1016/j.techfore.2010.09.006
http://dx.doi.org/10.1371/journal.pbio.1001109
http://dx.doi.org/10.1371/journal.pbio.1001109
http://www.nickbostrom.com/old/predict.html
http://www.nickbostrom.com/old/predict.html
http://www.jetpress.org/volume9/risks.html
http://dx.doi.org/10.1016/S0016-3287(03)00026-0
http://dx.doi.org/10.1016/S0016-3287(03)00026-0
http://dx.doi.org/10.1109/MC.1987.1663532

Minds
Bugaj, Stephan Vladimir, and Ben Goertzel. 2007. “Five Ethical
Imperatives and Their Implications for
Human-AGI Interaction.” Dynamical Psychology. http : / /
goertzel . org / dynapsyc / 2007 /
Five_Ethical_Imperatives_svbedit.htm.
Chalmers, David John. 2010. “The Singularity: A Philosophical
Analysis.” Journal of Consciousness Studies
17 (9–10): 7–65.
http://www.ingentaconnect.com/content/imp/jcs/2010/00000017
/
f0020009/art00001.
Cormen, Thomas H., Charles E. Leiserson, Ronald L. Rivest,
and Clifford Stein. 2009. Introduction to
Algorithms. 3rd ed. Cambridge, MA: MIT Press.
Cosmides, Leda, and John Tooby. 1994. “Beyond Intuition and
Instinct Blindness: Toward an Evolu-
tionarily Rigorous Cognitive Science.” Cognition 50 (1–3): 41–
77. doi:10.1016/0010- 0277(94)
90020-5.
Deaner, Robert O., Karin Isler, Judith Burkart, and Carel van
Schaik. 2007. “Overall Brain Size, and Not
Encephalization Quotient, Best Predicts Cognitive Ability
Across Non-Human Primates.” Brain,
Behavior and Evolution 70 (2): 115–124.
doi:10.1159/000102973.
Esser, James K. 1998. “Alive and Well After 25 Years: A
Review of Groupthink Research.” Organizational
Behavior and Human Decision Processes 73 (2–3): 116–141.
doi:10.1006/obhd.1998.2758.

[email protected] 2012. “Client Statistics by OS.” Stanford
University. Accessed January 2012. http :
//fah-web.stanford.edu/cgi-bin/main.py?qtype=osstats.
Gigerenzer, Gerd, and Henry Brighton. 2009. “Homo
Heuristicus: Why Biased Minds Make Better
Inferences.” Topics in Cognitive Science 1 (1): 107–143.
doi:10.1111/j.1756-8765.2008.01006.
x.
Gustafson, John L. 1988. “Reevaluating Amdahl’s Law.”
Communications of the ACM 31 (5): 532–533.
doi:10.1145/42411.42415.
Hall, John Storrs. 2007. Beyond AI: Creating the Conscience of
the Machine. Amherst, NY: Prometheus
Books.
. 2008. “Engineering Utopia.” In Wang, Goertzel, and Franklin
2008, 460–467.
Hanson, Robin. 1994. “If Uploads Come First: The Crack of a
Future Dawn.” Extropy 6 (2). http :
//hanson.gmu.edu/uploads.html.
. 1998. “Economic Growth Given Machine Intelligence.”
Unpublished manuscript. Accessed
May 15, 2013. http://hanson.gmu.edu/aigrow.pdf.
. 2008. “Economics of the Singularity.” IEEE Spectrum 45 (6):
45–50. doi:10 . 1109 / MSPEC .
2008.4531461.
Honkela, Timo, Ville Könönen, Tiina Lindh-Knuutila, and
Mari-Sanna Paukkeri. 2008. “Simulating

Processes of Concept Formation and Communication.” Journal
of Economic Methodology 15 (3): 245–
259. doi:10.1080/13501780802321350.
Kaas, Steven, Steve Rayhawk, Anna Salamon, and Peter
Salamon. 2010. Economic Implications of Software
Minds. The Singularity Institute, San Francisco, CA, August 10.
http://intelligence.org/
files/EconomicImplications.pdf.
Kahneman, Daniel. 2011. Thinking, Fast and Slow. 1st ed. New
York: Farrar, Straus / Giroux.
Kurzban, Robert. 2010. Why Everyone (Else) Is a Hypocrite:
Evolution and the Modular Mind. Princeton,
NJ: Princeton University Press.
16
http://goertzel.org/dynapsyc/2007/Five_Ethical_Imperatives_sv
bedit.htm
http://goertzel.org/dynapsyc/2007/Five_Ethical_Imperatives_sv
bedit.htm
/f0020009/art00001
/f0020009/art00001
http://dx.doi.org/10.1016/0010-0277(94)90020-5
http://dx.doi.org/10.1016/0010-0277(94)90020-5
http://dx.doi.org/10.1159/000102973
http://dx.doi.org/10.1006/obhd.1998.2758
http://fah-web.stanford.edu/cgi-bin/main.py?qtype=osstats
http://fah-web.stanford.edu/cgi-bin/main.py?qtype=osstats
http://dx.doi.org/10.1111/j.1756-8765.2008.01006.x
http://dx.doi.org/10.1111/j.1756-8765.2008.01006.x
http://dx.doi.org/10.1145/42411.42415

http://hanson.gmu.edu/uploads.html
http://hanson.gmu.edu/uploads.html
http://hanson.gmu.edu/aigrow.pdf
http://dx.doi.org/10.1109/MSPEC.2008.4531461
http://dx.doi.org/10.1109/MSPEC.2008.4531461
http://intelligence.org/files/EconomicImplications.pdf
http://intelligence.org/files/EconomicImplications.pdf
Kaj Sotala
Kurzban, Robert, and C. Athena Aktipis. 2007. “Modularity and
the Social Mind: Are Psycholo-
gists Too Self-Ish?” Personality and Social Psychology Review
11 (2): 131–149. doi:10 . 1177 /
1088868306294906.
Legg, Shane, and Marcus Hutter. 2007. “Universal Intelligence:
A Definition of Machine Intelligence.”
Minds and Machines 17 (4): 391–444. doi:10.1007/s11023-007-
9079-x.
Lino, Mark. 2010. Expenditures on Children by Families, 2009.
Miscellaneous Publication, 1528-2009.
Center for Nutrition Policy / Promotion, US Department of
Agriculture, June.
Lloyd, Seth. 2000. “Ultimate Physical Limits to Computation.”
Nature 406 (6799): 1047–1054. doi:10.
1038/35023282.
Loewenstein, George. 2005. “Hot-Cold Empathy Gaps and
Medical Decision Making.” Supplement,
Health Psychology 24 (4): S49–S56. doi:10.1037/0278-
6133.24.4.S49.

Loosemore, Richard. 2007. “Complex Systems, Artificial
Intelligence and Theoretical Psychology.” In Ad-
vances in Artificial General Intelligence: Concepts,
Architectures and Algorithms—Proceedings of the AGI
Workshop 2006, edited by Ben Goertzel and Pei Wang, 159–
173. Frontiers in Artificial Intelligence
and Applications 157. Amsterdam: IOS.
Mainzer, Klaus, ed. 2010. ECAP10: VIII European Conference
on Computing and Philosophy. Munich: Dr.
Hut.
McDaniel, Michael A. 2005. “Big-Brained People are Smarter:
A Meta-Analysis of the Relationship
between In Vivo Brain Volume and Intelligence.” Intelligence
33 (4): 337–346. doi:10 . 1016 / j .
intell.2004.11.005.
Moravec, Hans P. 1998. “When Will Computer Hardware Match
the Human Brain?” Journal of Evolution
and Technology 1.
http://www.transhumanist.com/volume1/moravec.htm.
Muehlhauser, Luke, and Louie Helm. 2012. “The Singularity
and Machine Ethics.” In Singularity Hy-
potheses: A Scientific and Philosophical Assessment, edited by
Amnon Eden, Johnny Søraker, James H.
Moor, and Eric Steinhart. The Frontiers Collection. Berlin:
Springer.
Mueller, Dennis C. 2003. Public Choice III. 3rd ed. New York:
Cambridge University Press.
Olson, Mancur, Jr. 1965. The Logic of Collective Action: Public
Goods and the Theory of Groups. Harvard

Economic Studies 124. Cambridge, MA: Harvard University
Press.
Omohundro, Stephen M. 2008. “The Basic AI Drives.” In Wang,
Goertzel, and Franklin 2008, 483–492.
President’s Council of Advisors on Science and Technology.
2010. Designing a Digital Future: Federally
Funded Research and Development in Networking and
Information Technology. Report to the President
and Congress. Executive Office of the President, Washington,
DC. http://www.whitehouse.
gov/sites/default/files/microsites/ostp/pcast-nitrd-report-
2010.pdf.
Rajab, Moheeb Abu, Jay Zarfoss, Fabian Monrose, and Andreas
Terzis. 2007. “My Botnet is Bigger
than Yours (Maybe, Better than Yours): Why Size Estimates
Remain Challenging.” In Proceedings
of 1st Workshop on Hot Topics in Understanding Botnets
(HotBots ’07). Berkeley, CA: USENIX. http:
//static.usenix.org/event/hotbots07/tech/full_papers/rajab/rajab.
pdf.
Salamon, Anna, Steve Rayhawk, and János Kramár. 2010. “How
Intelligible is Intelligence?” In Mainzer
2010.
Sandberg, Anders, and Nick Bostrom. 2008. Whole Brain
Emulation: A Roadmap. Technical Report,
2008-3. Future of Humanity Institute, University of Oxford.
http://www.fhi.ox.ac.uk/wp-
content/uploads/brain-emulation-roadmap-report1.pdf.
17

http://dx.doi.org/10.1007/s11023-007-9079-x
http://dx.doi.org/10.1037/0278-6133.24.4.S49
http://dx.doi.org/10.1016/j.intell.2004.11.005
http://dx.doi.org/10.1016/j.intell.2004.11.005
http://www.transhumanist.com/volume1/moravec.htm
http://www.whitehouse.gov/sites/default/files/microsites/ostp/pc
ast-nitrd-report-2010.pdf
http://www.whitehouse.gov/sites/default/files/microsites/ostp/pc
ast-nitrd-report-2010.pdf
http://static.usenix.org/event/hotbots07/tech/full_papers/rajab/ra
jab.pdf
http://static.usenix.org/event/hotbots07/tech/full_papers/rajab/ra
jab.pdf
http://www.fhi.ox.ac.uk/wp-content/uploads/brain-emulation-
roadmap-report1.pdf
http://www.fhi.ox.ac.uk/wp-content/uploads/brain-emulation-
roadmap-report1.pdf
Minds
Sandberg, Anders, and Nick Bostrom. 2011. Machine
Intelligence Survey. Technical Report, 2011-1. Fu-
ture of Humanity Institute, University of Oxford.
www.fhi.ox.ac.uk/reports/2011-1.pdf.
Shulman, Carl. 2010. Whole Brain Emulation and the Evolution
of Superorganisms. The Singularity Insti-
tute, San Francisco, CA. http://intelligence.org/files/WBE-
Superorgs.pdf.

Shulman, Carl, and Anders Sandberg. 2010. “Implications of a
Software-Limited Singularity.” In
Mainzer 2010.
Solomonoff, Ray J. 1985. “The Time Scale of Artificial
Intelligence: Reflections on Social Effects.” Human
Systems Management 5:149–153.
Sotala, Kaj, and Harri Valpola. 2012. “Coalescing Minds: Brain
Uploading-Related Group Mind
Scenarios.” International Journal of Machine Consciousness 4
(1): 293–312. doi:10 . 1142 /
S1793843012400173.
Stanovich, Keith E. 2008. What Intelligence Tests Miss: The
Psychology of Rational Thought. New Haven,
CT: Yale University Press.
Stanovich, Keith E., and Richard F. West. 1998. “Individual
Differences in Rational Thought.” Journal of
Experimental Psychology: General 127 (2): 161–188.
doi:10.1037/0096-3445.127.2.161.
Tooby, John, and Leda Cosmides. 1992. “The Psychological
Foundations of Culture.” In The Adapted
Mind: Evolutionary Psychology and the Generation of Culture,
edited by Jerome H. Barkow, Leda Cos-
mides, and John Tooby, 19–136. New York: Oxford University
Press.
Trivers, Robert L. 2000. “The Elements of a Scientific Theory
of Self-Deception.” Annals of the New York
Academy of Sciences 907 (1): 114–131. doi:10.1111/j.1749-
6632.2000.tb06619.x.
Uddin, Lucina Q., Marco Iacoboni, Claudia Lange, and Julian

Paul Keenan. 2007. “The Self and Social
Cognition: The Role of Cortical Midline Structures and Mirror
Neurons.” TRENDS in Cognitive
sciences 11 (4): 153–157. doi:10.1016/j.tics.2007.01.001.
Vinge, Vernor. 1993. “The Coming Technological Singularity:
How to Survive in the Post-Human Era.”
In Vision-21: Interdisciplinary Science and Engineering in the
Era of Cyberspace, 11–22. NASA Con-
ference Publication 10129. NASA Lewis Research Center.
http://ntrs.nasa.gov/archive/
nasa/casi.ntrs.nasa.gov/19940022855_1994022855.pdf.
. 2008. “Signs of the Singularity.” IEEE Spectrum, June. http : /
/ spectrum . ieee . org /
biomedical/ethics/signs-of-the-singularity.
Wang, Pei, Ben Goertzel, and Stan Franklin, eds. 2008.
Artificial General Intelligence 2008: Proceedings of
the First AGI Conference. Frontiers in Artificial Intelligence
and Applications 171. Amsterdam: IOS.
Yudkowsky, Eliezer. 1996. “Staring into the Singularity.”
Unpublished manuscript. Last revised May 27,
2001. http://yudkowsky.net/obsolete/singularity.html.
. 2001. Creating Friendly AI 1.0: The Analysis and Design of
Benevolent Goal Architectures. The Sin-
gularity Institute, San Francisco, CA, June 15.
http://intelligence.org/files/CFAI.pdf.
. 2007. “Levels of Organization in General Intelligence.” In
Artificial General Intelligence, edited
by Ben Goertzel and Cassio Pennachin, 389–501. Cognitive
Technologies. Berlin: Springer. doi:10.
1007/978-3-540-68677-4_12.

. 2008a. “Artificial Intelligence as a Positive and Negative
Factor in Global Risk.” In Global Catas-
trophic Risks, edited by Nick Bostrom and Milan M. Ćirković,
308–345. New York: Oxford Univer-
sity Press.
18
www.fhi.ox.ac.uk/reports/2011-1.pdf
http://intelligence.org/files/WBE-Superorgs.pdf
http://dx.doi.org/10.1037/0096-3445.127.2.161
http://dx.doi.org/10.1111/j.1749-6632.2000.tb06619.x
http://dx.doi.org/10.1016/j.tics.2007.01.001
http://ntrs.nasa.gov/archive/nasa/casi.ntrs.nasa.gov/1994002285
5_1994022855.pdf
http://ntrs.nasa.gov/archive/nasa/casi.ntrs.nasa.gov/1994002285
5_1994022855.pdf
http://spectrum.ieee.org/biomedical/ethics/signs-of-the-
singularity
http://spectrum.ieee.org/biomedical/ethics/signs-of-the-
singularity
http://yudkowsky.net/obsolete/singularity.html
http://intelligence.org/files/CFAI.pdf
http://dx.doi.org/10.1007/978-3-540-68677-4_12
http://dx.doi.org/10.1007/978-3-540-68677-4_12
Kaj Sotala
. 2008b. “Hard Takeoff.” Less Wrong (blog), December 2.
http://lesswrong.com/lw/wf/
hard_takeoff/.

19
http://lesswrong.com/lw/wf/hard_takeoff/
http://lesswrong.com/lw/wf/hard_takeoff/Abstract1
Introduction1.1 Intelligence, Optimization Power, and
Advantages2 Hardware Advantages2.1 Superior Processing
Power2.1.1 Superior Serial Power2.1.2 Increased Parallel
Power, Increased Memory3 Self-Improvement Advantages3.1
Improving Algorithms3.2 Designing New Mental Modules3.3
Modifiable Motivation Systems4 Co-operative Advantages4.1
Copyability4.2 Perfect Co-operation4.3 Superior
Communication4.4 Transfer of Skills5 Human Handicaps5.1
Biases from Computational Limitations or False
Assumptions5.2 Human-Centric Biases5.3 Biases from Socially
Motivated Cognition6 Discussion7 ConclusionsReferences

Running head ARTIFICIAL INTELLIGENCE1ARTIFICIAL INTELLIGENCE.docx

More Related Content

Similar to Running head ARTIFICIAL INTELLIGENCE1ARTIFICIAL INTELLIGENCE.docx (20)

More from toddr4 (20)

Recently uploaded (20)

Running head ARTIFICIAL INTELLIGENCE1ARTIFICIAL INTELLIGENCE.docx