ArticlePDF Available

Epistemology of AI Revisited in the Light of the Philosophy of Information



Artificial intelligence has often been seen as an attempt to reduce the natural mind to informational processes and, consequently, to naturalize philosophy. The many criticisms that were addressed to the so-called “old-fashioned AI” do not concern this attempt itself, but the methods it used, especially the reduction of the mind to a symbolic level of abstraction, which has often appeared to be inadequate to capture the richness of our mental activity. As a consequence, there were many efforts to evacuate the semantical models in favor of elementary physiological mechanisms simulated by information processes. However, these views, and the subsequent criticisms against artificial intelligence that they contain, miss the very nature of artificial intelligence, which is not reducible to a “science of the nature”, but which directly impacts our culture. More precisely, they lead to evacuate the role of the semantic information. In other words, they tend to throw the baby out with the bath-water. This paper tries to revisit the epistemology of artificial intelligence in the light of the opposition between the “sciences of nature” and the “sciences of culture”, which has been introduced by German neo-Kantian philosophers. It then shows how this epistemological view opens on the many contemporary applications of artificial intelligence that have already transformed—and will continue to transform—all our cultural activities and our world. Lastly, it places those perspectives in the context of the philosophy of information and more particularly it emphasizes the role played by the notions of context and level of abstraction in artificial intelligence. KeywordsArtificial Intelligence-Epistemology-Philosophy of information-Humanities-Sciences of the nature
Noname manuscript No.
(will be inserted by the editor)
Epistemology of AI Revisited in the Light of the
Philosophy of Information
Jean-Gabriel Ganascia
Received: date / Accepted: date
Abstract Artificial intelligence has often been seen as an attempt to reduce the nat-
ural mind to informational processes and, consequently, to naturalize philosophy. The
many criticisms that were addressed to the so-called “old-fashioned AI” do not concern
this attempt itself, but the methods it used, especially the reduction of the mind to a
symbolic level of abstraction, which has often appeared to be inadequate to represent
the richness of our mental activity. As a consequence, there were many efforts to evac-
uate the semantical models in favor of elementary physiological mechanisms simulated
by information processes. However, these views, and the subsequent criticisms against
artificial intelligence that they contain, miss the very nature of artificial intelligence,
which is not reducible to a “science of the nature”, but which directly impacts our
culture. More precisely, they lead to evacuate the role of the semantic information.
In other words, they tend to throw the baby out with the bath-water. This paper
tries to revisit the epistemology of artificial intelligence in the light of the opposition
between the “sciences of nature” and the “sciences of culture”, which have been intro-
duced by German neo-Kantian philosophers. It then shows how this epistemological
view opens on the many contemporary applications of artificial intelligence that have
already transformed – and will continue to transform – all our cultural activities and
our world. Lastly, it places those perspectives in the context of the philosophy of infor-
mation and more particularly it emphasizes the role played by the notions of context
and level of abstraction in artificial intelligence.
Keywords Artificial Intelligence ·Epistemology ·Philosophy of Information ·
Humanities ·Sciences of the Nature
Jean-Gabriel Ganascia
LIP6, University Pierre and Marie Curie (Paris VI)
4, place Jussieu, 75252, Paris, Cedex 05, France
Tel.: +33 (0) 1 44 27 37 27
Fax: +33 (0) 1 44 27 70 00
1 Introduction
Traditional Artificial Intelligence (AI) has often been accused of failing to deliver on
its promises. Some people currently say that there is something wrong in what they
condescendingly call the “Good Old-Fashioned” symbolic AI that has been accused
of oversimplifying the world. More precisely, it has been said that the reproduction of
high level cognitive abilities, for instance mathematics, reasoning or playing chess, were
easier, from a computational point of view, but less valuable than the simulation of basic
physiological mechanisms of perception and action. Frequently invoked by specialists of
robotics and AI in the last 20 years, “Moravec’s Paradox” (Moravec, 1988) summarizes
this point: it claims that cognitive abilities, which require ratiocination, are easier to
simulate on a computer with a few logical rules than low level cognitive abilities, like
perception. For instance, nowadays, there are artificial intelligence programs playing
chess, proving theorems or interpreting natural language queries. However, “low level”
cognitive abilities, such as the ability to recognize faces or to clean the dishes, seem
to be much more difficult to implement than those intellectual faculties. Similarly,
basic animal behaviors, e.g. the capacity to perceive or awareness, seem very difficult
to reproduce using logical and deterministic mechanisms. This is paradoxical, because
the higher intelligence activities, which are proper to humans, seem to be easier to
reproduce with classical AI techniques than the basic physiological mechanisms that
almost all species possess.
As a consequence, many of those who mock and criticize traditional AI, which is
restricted to the simulation of high level cognitive abilities, promote what they call
a “Nouvelle AI” that would effectively mimic physiological processes. More generally,
they propose to build efficient machines by increasing the complexity of the models
and by designing powerful mechanisms that reproduce basic animal capacities (Brooks,
Obviously, it is very difficult to conclude the effective failure of AI, because, as we
shall see further on in this article, compared to many other contemporary disciplines, AI
has achieved many successes. Careful attention to the criticisms addressed to traditional
AI shows that this disapproval is not really caused by so-called failures or by a non
fruitfulness of AI, but by a philosophical divergence that implies a difference of attitude
towards intelligence and science. Symmetrically, one can note that, since its initial
promotion, more than twenty-five years ago, in the mid-eighties, the “Nouvelle AI”
did not so much contribute to successful achievements, which would justify its claims
unquestionably. Lastly, the conceptual basis of the “Nouvelle AI”, which is rooted on
cybernetics and dynamic system theory, are older than those of traditional Artificial
It therefore would seem that many of the accusations against AI are due to a
misunderstanding of its project and are not caused by intrinsic weaknesses of AI tech-
niques. Behind this misunderstanding, there is confusion on the epistemological status
of Artificial Intelligence. This article constitutes an attempt to elucidate those points.
It first shows, as we have just suggested, that most of the critics that were traditionally
addressed against symbolic AI aren’t justified. It then tries to elucidate the philosoph-
ical status of traditional AI by reference to both the pioneers of artificial intelligence
and some philosophical works, in particular to Floridi’s Philosophy of Information (PI)
(Floridi, 2010) and to the traditional distinction between the “Sciences of Nature” and
the “Sciences of Culture” that was introduced by the Neo-Kantian school of philosophy.
Lastly, it demonstrates how it is different from both the “Nouvelle AI” philosophical
status, which is no more than a promotion of an old materialist philosophy, and first
2 The Paradoxical Failure of AI
As previously noted, it is a commonplace nowadays to say that the “Good old-fashioned
AI” has failed to deliver its promises. And, there is nothing more characteristic of a
philosophical attitude than to analyze and to discuss commonplaces. Nevertheless, in
the present case, it’s risky, since this commonplace is admitted by almost everybody
without any discussion. Moreover, it could be wrongly interpreted and viewed as a
defense of a corporation, while it is the philosophical scope of those criticisms that is
of interest here. Our purpose is to show that there is some confusion behind the way
AI is understood. More precisely, depending on the philosophical perspective that is
adopted, AI may be envisaged differently. But, before investigating this point in more
depth, let us recall the numerous successes of AI.
2.1 The Successes of AI
Let us first recall that AI is an academic discipline that was born in 1956 during
the Dartmouth Summer Research Conference on Artificial Intelligence. This event was
organized by a young logician who was less than thirty years old, John McCarthy.
The proposal was signed by four persons: John McCarthy, Marvin Minsky, Nathaniel
Rochester and Claude Shannon. It was explicitly based on “the conjecture that every
aspect of learning or any other feature of intelligence can in principle be so precisely
described that a machine can be made to simulate it.” (McCarthy et al., 1955). Accord-
ingly, the original AI project was not to reify an intelligent being with a machine, but
to decompose intelligence into numerous features and to simulate each of them with a
computer. Note that, despite the fact that John McCarthy, the organizer of the Dart-
mouth Summer Research Conference on AI, was a logician, the proposed techniques to
achieve this objective were not at all restricted to symbolic computation. For instance,
in the McCarthy proposition, it was said (cf. (McCarthy et al., 1955)) that among the
different aspects of the AI problem included:
1. Automatic Computers
2. How Can a Computer be Programmed to Use a Language
3. Neuron Nets
4. Theory of the Size of a Calculation
5. Self-Improvement
6. Abstractions
7. Randomness and Creativity
Now, let us examine some of the AI achievements more than fifty years after its birth.
There were many successful attempts to reproduce the ratiocination. For instance Au-
tomatic Theorem Proving has been so well developed that now many mathematical
activities have evolved, due to the introduction of computers that partially automate
the proof or the proof checking. It even appears that Mathematical Logic has been
strongly influenced by the development of artificial intelligence: new research areas
have been elaborated to answer to the AI needs, for instance the Non-Monotonic Log-
ics, the Description Logics or the numerous Logics for Agents. And, the reproduction
of ratiocination is not restricted to mathematics: everyone knows that nowadays world
chess champions are all defeated by machine. However, AI is not restricted to abstract
logical and mathematical reasoning. Natural Language Processing has been intensively
developed with many successful applications. It is not only to make the machine un-
derstand or translate natural language texts, but also to tag parts of speech, to extract
meaningful patterns and to improve our knowledge of many syntactic and semantic
phenomenon. Note that the extensive use of Data Mining and Knowledge Extraction
has transformed the corpus linguistic, making possible the process of enormous quanti-
ties of texts. Perception has also been simulated with captors and pattern recognition
techniques: it is now possible to understand speech, to identify visual patterns and to
design robots able to perceive and to navigate using their own representation of the
AI techniques also play a key role in the simulation of memory. Let us recall that
the invention of the hypertext by Ted Nelson in 1965, which was designed as an ex-
ternal memory in reference to Vanevar Bush’s MEMEX, was directly influenced by
the development of AI techniques, in particular by list processing. It explains the ti-
tle of Ted Nelson’s seminal paper (Nelson, 1965), A File Structure for The Complex,
The Changing and Indeterminate. Later on, in the seventies, the modeling of seman-
tic memory, e.g. semantic networks (Quillian, 1968), took advantage of AI techniques.
In return, semantics helped to design new knowledge representation techniques, e.g.
frames (Minsky, 1975), which constitutes the semantic turn of AI. More recently, the
web was designed by Tim Berners-Lee as a model of memory and then the Semantic
Web (Berners-Lee et al., 2001) made an extensive use of AI Knowledge Representation
techniques. For instance, the notion of Resource Description Framework Schemas is
nothing more than semantic networks applied to describe meta-data of web resources.
From a practical point of view, AI greatly contributed to contemporary information
society and, consequently to what Luciano Floridi calls the Fourth Revolution (Floridi,
2008, 2010). Let us take two examples. The first concerns object oriented languages,
which are nowadays the most frequently used programming languages: they are directly
inspired from AI semantic knowledge representation techniques. The second concerns
the web. As previously said, it is a model of memory built on hypertext links, which
are directly influenced by AI. And more recently the evolution of the web towards the
Semantic Web using ontologies,description logics and other Knowledge Representation
techniques shows how strong is the AI influence. In the same way, the notion of ambient
intelligence refers also to AI.
As a conclusion, it is difficult to imagine a discipline whose concepts so strongly
contributed to the evolution of society in less than half a century. Therefore, the men-
tion of a so-called failure of AI and the condescending manner in which one refers to
“Good Old-Fashioned AI” are a little bit paradoxical.
2.2 The Misunderstanding
We would like to understand the paradox according to which so many people argue
against AI, while, as previously said, this academic discipline was both so successful
and so influential in the contemporary information society. Undoubtedly, AI is a very
fascinating project, which makes people simultaneously very excited and suspicious.
But, while this project is very ambitious, it is also ambiguous. Depending of the sig-
nificance of the locution Artificial Intelligence, its aim can be interpreted differently:
it is either to build an intelligent being using information technologies or to decom-
pose intelligence and to describe precisely each of its components in a way that can be
simulated on a computer.
In its first significance AI is understood as an attempt to reproduce consciousness
or, at least, to reify the mind with a machine. This first significance is widespread and
very popular. It made AI enjoyable. From a philosophical point of view, it means that
our consciousness – or our mind – is nothing more than a sum of mechanisms, which
satisfies most of our contemporaries. But, this significance is perforce very disappoint-
ing, because it proposes to produce a machine that owns a kind of consciousness, or at
least of mind, which is too ambitious to be achieved in the short term. Even if the mind
or the consciousness can, in principle, be simulated with a the machine, this simulation
constitutes a horizon of possibility that is not attainable very soon. As a consequence,
AI inevitably fails to fulfill the hopes of impatient persons. Regularly, people argue
against AI saying that AI is not able to attain its objectives. Some say that the objec-
tives of AI are not attainable; others say that the methods of AI are not appropriate.
The so-called “Nouvelle AI” explains that the failure is caused by the restriction of
the mind to symbolic manipulation, without paying attention to physiology and to the
body part of intelligence.
In its second significance, AI is a scientific discipline that studies how intelligence
can be so precisely decomposed in its different aspects that each of them can be re-
produced on computers. This meaning corresponds to the definition given by John
McCarthy in the above mentioned project for the Dartmouth Summer Research Con-
ference in 1956 (cf. (McCarthy et al., 1955)). According to this second significance,
AI investigates intelligence with resources of the artificial (cf. (Simon, 1996)), i.e. with
information processing techniques, without having to produce an artificial mind. So
doing, AI is what Herbert Simon named a Science of the Artificial; it is not restricted
to what Allen Newell described as Physical Symbol Systems (Newell, 1980) and to
psychological simulation; there may be many other levels of abstraction that can be
simulated using AI techniques. In this respect, it is in total accordance with the prin-
ciple of the Floridi’s Philosophy of Information (Floridi, 2010). Note, for instance,
that Neural Nets were explicitly mentioned in the McCarthy project (McCarthy et al.,
1955), which means that AI is not restricted to a naive Cartesian dualism, as some
detractors want to let believe.
Clearly, AI was very successful and the results that were obtained greatly con-
tributed to the fast development of information technologies. And, this discipline is
always active and continues to produce new scientific results that will be useful in the
future. However, understood as a scientific discipline that attempts to elucidate and
simulate the intelligence with machines, AI is far less fascinating in its first significance.
This explains why the debate about AI is mainly focused on the first significance, which
is a pity, because many questions that are of interest from a philosophical point of view,
and especially from the point of view of the Philosophy of Information, concern the
possibility and the methods of AI understood as a scientific discipline. It does not mean
that the discussion about the reduction and/or the reproduction of the mind and/or
the consciousness with computers are not justified. However, even if these issues are
more commonly discussed, they are far less current, because the practical results of
AI understood as a Science of the Artificial strongly contributes to transform our con-
temporary world, while there is no evidence that it is already possible to very soon
reproduce a mind or a consciousness with a computer, or with any other contemporary
3 What Went Wrong in AI?
Two years ago, a special issue of AI Magazine (Shapiro and Goker, 2008) published
numerous of cases of alleged faulty AI systems. The goal was to understand what
made them wrong. The main lesson was that, most of the time, the difficulties were
not due to technical impediments, but to the social inadequacy of the AI systems to
their environment. This point is crucial. It has motivated the reflexion presented in
this paper.
3.1 Elves Keep You
For the sake of clarity, let us take an example about the so-called electronic “elves”,
which are personal agents who act as efficient secretaries and help individuals to manage
their diary, to fix appointments, to find rooms for meetings, to organize travel, etc. A
paper (Knoblock et al., 2008) published in the above mentioned special issue of the
AI Magazine (Shapiro and Goker, 2008) reported technical successes but difficulties
with some inappropriate agent behaviors. For instance, one day, or rather one night,
an elf rang his master at 3 am to inform him that his 11 o’clock plane was going to
be delayed. Another was unable to understand that his master was not available for
anybody in his office, since he had to complete an important project... Many of these
inappropriate actions make those intelligent agents tiresome and a real nuisance, which
causes their rejection by users.
3.2 An Embarrassing Investment Adviser
A few years ago, I was a consultant for a large French bank. The management wanted
to introduce knowledge technologies in the company’s culture. The reason was that
the managers complained that in bank agencies, people in charge of helping customers
were unable to provide relevant expert advices because they were only familiar with
two or three products among the full range of available solutions. As a consequence,
they systematically advised the products they knew, forgetting the others, even when
they were more appropriate. The managers thought that a knowledge-based system
could advantageously replace – or possibly train – those poor investment advisers.
This is why they got in touch with my group who they asked to build a Knowledge
Based-System able to act as an efficient adviser that helps customers to invest their
My group succeeded in building an efficient “investment adviser” by using the
knowledge engineering techniques that were in used at that time. The resulting system
asked relevant queries, diagnosed the situation of the customers and provided, for each
of them, eligible, diversified and judicious investments that take advantage of all the
products proposed by the bank. From a technical point of view, it seemed that it gave
entire satisfaction.
However, the system has never been in use for two reasons. The first was the
refusal from the bank agency managers: they feared being reduced to a simple role of
The second came from customers who suspected the AI systems were provided
by the bank to serve the interests of the bank. Note that surprisingly they were not
so much suspicious of the bank employees nor the bank softwares, but rather the AI
systems provided by the bank.
3.3 The Social Dimension of AI
Those two examples show that social inadequacy is the main cause of AI system rejec-
tion. In both cases, the AI programs were technically successful; they were not accepted
because they did not answer to the requirements of the social environment where they
had to be used. The causes of inappropriateness were not in the artificial system itself,
but in the adequacy of the artificial system to the surroundings.
This conclusion is neither astonishing nor original. Many people have noticed that
the failures of knowledge-based systems were mainly due to man-machine interfaces or
to organizational impediments, which made them inefficient (cf. for instance (Hatchuel
and Weil, 1995)). Moreover, it is in accordance with what the pioneers of AI had said,
and in particular Herbert Simon who has insisted on the importance of the outer en-
vironment in his famous book “The Science of the Artificial” (Simon, 1996): according
to him, “Human beings, viewed as behaving systems, are quite simple. The apparent
complexity of our behavior over time is largely a reflection of the complexity of the
environment in which we find ourselves.” In other words, the difficulty would not be
in reproducing intelligent behaviors, but in adapting them to the complexity of their
3.4 Why Would Something be Wrong?
These conclusions are so obvious and conform with the predictions that the above
mentioned AI failures would have had an incentive to address both user-centered design
and social studies. Nevertheless, surprisingly, since the eighties, the evolution of AI
toward, for instance, the so-called “Nouvelle AI” has gone in a completely different
direction: because it has been accused of oversimplifying the world and of ignoring the
physical bodies, AI has been tempted to increase the complexity of its models and to
build powerful machines able to effectively mimic physiological capacities.
This view tends to reduce AI to a simulation of the natural processes. It opens
undoubtedly exciting prospects for scientists. However, as we shall see in the following
section and as it was previously mentioned, this does not exhaust the project of AI,
which cannot be fully assimilated to a pure reproduction of the cognitive abilities,
i.e. to a “naturalization” of the mind. In other words, this project reduces AI to only
one of its significances, i.e. to the production of an artificial mind, or of an artificial
consciousness. However, it does not address the second significance of AI that is the
discipline that investigates intelligence by simulating it with artificial devices.
In a way, the opposition between those two significances of AI reflect an old oppo-
sition, introduced in the beginning of the 20th century, by neo-Kantian philosophers,
between the “Sciences of nature” and the “Sciences of culture”, i.e. the humanities.
Herbert Simon himself introduces the “Sciences of the artificial” to qualify the type
of investigation that motivated AI, which suggests exploring more in depth this tra-
ditional opposition to better understand the epistemological status of AI understood
as a scientific discipline and, subsequently the status of the Simon’s “Sciences of the
4 “Artificiality” vs. “Culturality”
4.1 The “Sciences of the Artificial”
Herbert Simon has introduced the distinction between the “Sciences of Nature” and
the “Sciences of the Artificial” in a famous essay published in 1962 in the “Proceedings
of the American Philosophy Society”. The question was of importance for him, since
he worked for more 35 years on it, he has re-edited the same book three times in 1969,
in 1980 and in 1996, and has considerably augmented the volume of the book: the first
edition published in 1969 contained 123 pages, while the third edition, published in
1996, contained 231 pages.
The original point of Herbert Simon was to introduce the notion of artificiality to
describe complex artificial systems in complex environments and to make them object
of science. According to him, artificial systems have to be distinguished from natural
systems, because they are produced by human beings – or, more generally, by intelligent
beings – who have in mind some goals to achieve. More precisely, artificial things are
characterized by the four following points (Simon, 1996):
1. They are produced by human (or by intelligent beings) activity.
2. They imitate more or less nature, while lacking the whole characteristics of natural
3. They can be characterized in terms of functions,goals and adaptation.
4. They can be discussed both in terms of imperatives or as descriptives.
Remark that the universe of artificial things is not reduced to the computerized world.
Many artificial objects that were invented far before the existence and the development
of electronic computers, for instance airplanes and clocks, own all the above mentioned
characteristics. However, computers greatly facilitate the building of artificial things.
Since artificial things can be approached not only in descriptive terms of their
structure, but with respect to their functions, their goals and their adaptive abilities,
they cannot be reduced to natural things that have only to be objectively described
from the outside, without any a priori. Their study can take into consideration the
imperatives to which they are supposed to obey. As a consequence, the discipline that
is in charge to study artificial things, i.e. the science of the artificial things, has to be
distinguished from the sciences of the natural things. To characterize this discipline,
Herbert Simon has introduced the concept of “artifact”, which is defined as an interface
between the “inner” environment, i.e. the internal environment of an agent, and the
“outer” environment where it is plunged. As previously said, the “inner” environment is
easy both to describe in terms of functions, goals and adaptation and to simulate with
computers; its complexity results from the “outer” environment in which it operates.
It has to be recalled that artificial things can always be studied with the methods
of the “sciences of nature”, for instance a clock can be studied from a physical point
of view, by analyzing the springs and the wheels it is composed of, but those “sciences
of nature” don’t take into consideration the imperatives to which artificial things are
supposed to obey, their functions and their goals.
Symmetrically, natural things can be investigated by the “sciences of the artificial”.
More precisely, according to Herbert Simon, the “sciences of the artificial” can greatly
help to improve our knowledge of the natural phenomenon. Any natural thing can be
approached by building models, i.e. artificial things, that aim at simulating some of
their functions. For instance, cognitive psychology has been very much improved by
the use of computers that help to simulate many of our cognitive abilities.
4.2 Limits of the Artificiality
Two criticisms can be addressed to AI understood as a “science of the artificial”.
The first is traditional and recurrent: for more than 20 years now, scientists and
philosophers criticize the oversimplified models of the so-called “old-fashioned AI”. In
a word, they think that models have to be exact images of what they are intended to
model. As a consequence, the “artifacts”, taken in Herbert Simon terms, i.e. the inter-
faces between “inner” and “outer” environments, have no real value when the “inner”
environments are too schematic. Therefore, the artificiality has to faithfully copy the
reality, i.e. nature. As a consequence, many mental and social phenomenon are viewed
as natural phenomenon. For instance, the mind is reduced to physical phenomenon that
result from brain activity (Manzotti, 2007) or the epistemology is identified to infor-
mational processes (Chaitin, 2006). The AI itself has been mathematized by physicists
as a unified and universal theory (Hutter, 2005), which gave birth to the General Arti-
ficial Intelligence. This tendency corresponds to the so-called “naturalization”, which
is very popular nowadays among philosophers (Dodig-Crnkovic, 2007). Nevertheless,
despite the huge amount of researches done in this area for many years now, only a
few results have been obtained.
The second criticism is symmetric: the notion of “artifact” does not allow to fully
approach the semantical and cultural nature of all mental processes. For instance,
Herbert Simon considers music as a science of the artificial, since everything that is
said about the sciences of the artificial can be said about music: it requires formal
structures and provokes emotions. It is partially true, however, music is not only a
syntax; semantical and cultural dimensions of music exist and they are not taken into
account in Simon models. Therefore, we pretend that an extension of the “science of
the artificial” toward the “sciences of culture” is required.
In other words, while the first criticism opens on a naturalization, i.e. on a refine-
ment of the models, the second pursues and extends the Herbert Simon “sciences of the
artificial” by reference to the Neo-Kantian “sciences of culture” that will be presented
in the next section.
5 The “Sciences of the Culture”
5.1 Origin of the “Sciences of the Culture”
The notion of “Sciences of the Culture” (Rickert, 1921) was introduced in the beginning
of the 20th century by a German Neo-Kantian philosopher, Heinrich Rickert who was
very influential on many people among which were the sociologist Max Weber and
the young Martin Heidegger. Its goal was to base the humanities, i.e. disciplines like
historic studies, sociology, laws, etc., on rigorous basis. More precisely, he wanted to
scientifically characterize the sense of human activities, i.e. culture understood as the
result of goal oriented activities. In other words, he wanted to build an empirical science
able to interpret human achievements as the results of mental processes. However,
he thought that the scientific characterization of the mind had to be distinguished
from the psychological science, i.e. from the psychology, which approached the mental
phenomenon with the methods of physical sciences. For him, spiritual phenomenon have
a specificity that cannot be reduced to a physical one, even if they can be submitted
to a rational and empirical inquiry. The distinction between “sciences of nature” and
“sciences of culture” had to precisely establish this specificity. As we shall see in the
following, according to Rickert, the underlying logic of the “sciences of culture” totally
differs from the logic of the “sciences of nature”.
Before going more into the detailed characterization of those approaches, let us add
that “sciences of culture” have nothing to do with “cultural studies”: the former at-
tempt to scientifically characterize the results of human conscious activities – politics,
art, religion, education, etc.– while the latter try to identify and to differentiate cul-
tural facts from various manifestations of human activities – dances, musics, writings,
sculpture, etc.–. Very often cultural studies aim at exploring the cultural specificities
and their conflict with official cultures and powers that tend to ignore them. As already
said, the notion of “sciences of culture” was introduced in the early 20th century, while
“cultural studies” only exist since the sixties. Lastly, “sciences of culture” do not pro-
mote culture as the expression of identities, while “cultural studies” are often advocate
of such expression.
As previously mentioned, the “sciences of culture” aim at understanding social phe-
nomenon that result from human conscious activities. Obviously, physics and chemistry
are out of the scope of the “sciences of culture” because they investigate the objec-
tive properties of the world, without any interference with human activities. On the
contrary, the study of religion and discrimination may participate to the “sciences of
the culture”. But, the distinction is not so much a difference in the objects of study
than in the methods of investigation. Therefore, the history of physics contributes to
the “sciences of culture” while some mathematical models of social phenomenon, e.g.
game theory, contribute to “sciences of nature”. Moreover, the same discipline may
simultaneously contributes to “sciences of nature” and to “sciences of culture”; it is
what Rickert characterizes as an intermediary domain. For instance, medicine bene-
fits simultaneously from large empirical studies and from individual case studies; the
former enter more likely into the logic of “sciences of the nature” and the latter into
the logic of “sciences of culture”. It even happens, in disciplines like medicine, that
national traditions differ, some of them being more influenced by the “sciences of na-
ture”, like evidence-based medicine, while others contribute more easily to the “sciences
of culture”, like clinical medicine when it is based on the study of the patient history.
In other words, the main distinction concerns different logics of sciences that are
described in the next section.
5.2 The Tree Logics
Ernst Cassirer clearly described the different logics of sciences in many of his essays
(Cassirer, 1923, 1961). Briefly speaking, he first distinguishes the theoretical sciences
like mathematics, which deal with abstract and perfect entities as numbers, figures of
functions, from empirical sciences that are confronted with the material reality of the
world. Then, among the empirical sciences, Ernst Cassirer differentiates “sciences of
nature”, which deal with physical perceptions, and “sciences of culture” that give sense
to the world. According to him and to Heinrich Rickert, “sciences of nature” proceed by
generalizing cases: they extract general properties of objects and they determine laws,
i.e. constant relations between observations. As a consequence, the logic of “sciences
of nature” is mainly inductive, even if the modalities of reasoning may be deductive or
abductive. The important point is that the particular cases have to be forgotten; they
have to be analyzed in general terms and composed of well defined objects that make
no reference to the context of the situation. The validity of the scientific activity relies
on the constance and the generality of the extracted laws.
By contrast to the logic of “sciences of nature”, the logic of the “sciences of culture”
do not proceed by generalizing multiple cases. It does not extract laws, i.e. relations
between observations; it does not even work with physical perceptions, but with mean-
ingful objects that have to be understood. In brief, the main function of “sciences of
culture” is to give sense to the world. Their way of investigation is to understand partic-
ulars. The general methodology is to observe individual cases and to understand them.
However, they have to choose, among the particulars, individuals that are paradig-
matic, i.e. who can teach general lessons that may be reused in other circumstances.
In other words, “sciences of culture” are not properly interested in the singularity of
cases, which has to be forgotten, but in the understandability of individuals under
study. Their methods help to give sense to observations of complex individual cases.
5.3 “Science of Culture” vs “Science of Artificial”
As previously said, culture can be understood as the result of goal oriented human
activities. For instance, agriculture is the art and practice of working soils to produce
crops and other vegetables. The “sciences of culture” try to understand the human
activities, i.e. the human goals and the ways humans take to reach them. Since AI tries
to reproduce intelligent human activities, it can obviously benefit from the methods of
the “sciences of culture”. However, it can also benefit from the theoretical sciences that
work on abstract entities, i.e. from mathematics and logic, and from the “sciences of
nature”, which, for instance, investigate physiological or physical mechanisms. Looking
back to the “sciences of the artificial”, it appears that they belong both to “sciences
of nature”, since they proceed by generalization of cases, and to “sciences of culture”,
because they characterize artificial things by their functions, their goals and their
adaptivity and not only by their structure.
The next section shows how methods of “sciences of culture” can play an important
role in AI, even if AI cannot be reduced to a “science of culture”. Nevertheless, the im-
portant point here concerns the distinction between the “sciences of the artificial” and
the “sciences of culture”. As previously said, the artificiality, taken in the sense given
by Herbert Simon, includes not only the things that are produced by the activity of in-
telligent beings, but also the goals to which they are designed for. Human productions
are not reducible to the material things they achieved. For instance, a statue is more
than the bronze it contains; a clock is more than the metal it is made of; a book is more
than paper and ink, etc. As a consequence, artificiality is also part of the culturality.
The sciences that produce artifacts, i.e. the “sciences of the artificial” are undoubtedly
part of “sciences of culture”, while culture covers a broader area since it also includes
pure interpretative activities like history. Moreover, the logic of “sciences of culture”
extends the logic used in the “sciences of the artificial” that remains partially similar
to the logic of “sciences of nature”.
6 AI as an intermediary domain
The thesis developed here is that the alleged AI weaknesses are not caused by the
oversimplification of AI models, as many people claim nowadays, but by their inade-
quacy to the “outer” environment. It has been shown that the notion of “science of the
artificial”, which was introduced by Herbert Simon, has to be extended by reference
to the notion of “science of culture”.
From a philosophical point of view, it means that AI participates to the “sciences
of culture”, i.e. that it cannot be entirely reducible to a “science of nature” or to
mathematics and theoretical sciences. But it is not more reducible to “sciences of
culture”. More precisely, it is what Heinrich Rickert identifies as an “intermediary
domain” that belongs simultaneously to theoretical sciences, i.e. to formal logic and
mathematics, to empirical sciences of nature and to empirical sciences of culture. The
practical consequences of such philosophical considerations are twofold: they have an
impact on both the methods and the objects of application of AI.
6.1 Methods of AI
Since AI contributes to “sciences of culture”, it has to take advantage of the logic of
“sciences of culture”, which may enlarge the scope of its methods. Let us recall that
“sciences of culture” are empirical sciences, i.e. they build knowledge from the obser-
vation of particulars. However, they don’t proceed by extracting properties common
to observed cases; they do not abstract knowledge from particulars. They collect data
about individual cases and they attempt to understand them, i.e. to find a common
cause or to give a reason for them. Let us specify that it is not to extract singularities,
but to investigate paradigmatic cases and to explain in what respect the individual
cases under study can be universalized.
An excellent example of such studies was done by a cognitive anthropologist, Ed-
win Hutchins, in the book titled “Cognition in the wild” (Hutchins, 1995) where he
attempted to identify the cognition in its natural habitat, in the circumstances a mod-
ern ship, and to model it. In practice, many preliminary studies should have recourse
to such methods. It has to be the case with knowledge engineering and, more generally,
when designing any AI concrete application.
Moreover, the attentive study of past failures contributes to this dimension of AI.
It is not to generalize all the individual failures by extracting their common properties,
as could be for any “science of nature”, but to understand the logic of the failures, as
did, for instance, Dietrich D¨orner in his book “The Logic of Failure”(D¨orner, 1997), to
see what lessons could be drawn from these bad experiences and to learn from them.
In this way, it could contribute to the logic of “sciences of culture”.
6.2 Objects of AI
Lastly, the investigations of AI could focus more deliberately on cultural dimensions of
the world, where there are many valuable applications. The information sciences and
technologies greatly contribute to the advancement of knowledge to the point where
the present age is often called the “knowledge age”. However, it’s a pity that AI did
not participate more actively in cultural evolutions consecutive to the development of
information technologies, for instance, to the Wikipedia free encyclopedia or to the
social web.
6.3 Perspectives for AI
More generally, the knowledge quest can be greatly accelerated by the use of AI tech-
nologies. For instance, my team is working in musicology (Rolland and Ganascia, 2002,
1999; Ramalho et al., 1999), in textual criticism (Bourdaillet et al., 2007), in social
sciences (Velcin and Ganascia, 2005), in epistemology (Ganascia and Debru, 2007;
Ganascia, 2008), in ethics (Ganascia, 2007) etc. But there are many other fields of
applications, not only in humanities. Let us insist that such applications of AI are
directly connected with cultural dimensions. So, in the case of medicine, there already
exist many attempts to model organs (Nobel, 2006) and to simulate medical diagnosis;
AI played a part in these successful achievements, which are related to “sciences of
nature”; but the new challenge now is to manage all the existing knowledge and to
help researchers to find their way. This is undoubtedly the role of AI understood as a
“science of culture” to help to achieve such tasks.
7 Conclusion and persp ective
To conclude, let us first insist on our main point: AI can neither be reduced to a “Science
of nature” nor to a “Science of culture”; it is what Rickert calls an “Intermediary
domain”. This has not only philosophical implications on the epistemological status
of AI, but also practical consequences about both the objects and the methods of AI.
Moreover, the reduction of AI to a “Science of nature” does not allow to understand the
role it plays in the development of our Information Society. The concept of knowledge
as it is commonly used today to qualify the present state of our societies does not only
refer to the democratization of education or to the high qualifications that are required
in a modern economy, but also to the formalization of interpretation processes, which
render possible the storage, the access and the exchange of knowledge. For instance, the
notion of ontology as it was developed in AI during the last few years takes its sense in
the context of a “Science of culture”, i.e. with respect to interpretation processes, but
not with respect to a “Science of nature”. To be convinced, let us quote Tom Gruber,
one of the most influential persons in the field of ontology design in AI, who said in
an interview that: “Every ontology is a treaty – a social agreement – among people
with some common motive in sharing.” (Gruber, 2004). More generally, the way AI
attributes meaning to symbols, i.e. the semantic in AI, does not refer to a “Science
of nature”, but to a “Science of culture”. To this respect, the notion of Knowledge
Level, which was introduced by Alan Newell in 1982 (Newell, 1982) and which was so
influential – and so controversial – in the field of Knowledge Acquisition in AI (Clancey,
1993) during the nineties, is illustrative: it does not reduce knowledge to symbols or
to information, but it makes knowledge the result of an interpretative process. More
generally, it refers knowledge to a specific Level of Abstraction which takes sense in
Everything which has been said here concerning AI is also valid for most of au-
tomatic information processes. As an illustration, the way semantic information is
extracted from data can neither be reduced to the sole induction, i.e. to a general-
ization from particulars, nor to a representation in a universal digital ontology. The
knowledge, which is relevant semantic information, takes its sense within interpretative
processes, at a Level of Abstraction and in a given Context, i.e. with respect to the key
concepts of the Philosophy of Information (Floridi, 2010). More generally, most of the
open problems of Philosophy of Information can be enlightened by being envisaged
under the light of the opposition between the “Sciences of nature” and the “Sciences
of culture”.
Berners-Lee, T., Hendler, J., and Lassila, O. (2001). The semantic web. Scientific
American Magazine.
Bourdaillet, J., Ganascia, J.-G., and F´enoglio, I. (2007). Machine assisted study of
writers’ rewriting processes. In Proceedings of the 4th International Workshop on
Natural Language Processing and Cognitive Science (NLPCS).
Brooks, R. (2002). Flesh and Machines: How Robots Will Change Us. Pantheon Books.
Cassirer, E. (1923). Substance and Function. Open Court, Chicago.
Cassirer, E. (1961). The Logic of the Humanities. Yale University Press, New Haven.
Chaitin, G. (2006). Epistemology as information theory. COLLAPSE, 1:27–51.
Clancey, W. (1993). The knowledge level reinterpreted. In Ford, K. M. and Bradshaw,
J. M., editors, Knowledge acquisition as modeling. John Wiley & Sons.
Dodig-Crnkovic, G. (2007). Epistemology naturalized: The info-computationalist ap-
proach. American Philosophy Association Newsletter on Philosophy and Computers,
orner, D. (1997). The Logic Of Failure: Recognizing And Avoiding Error In Complex
Situations. A Merloyd Lawrence Book, Perseus Books, Cambridge, Massachusetts.
Floridi, L. (2008). Artificial intelligence - new frontier: Artificial companions and the
fourth revolution. Metaphilosophy, 39(4/5):651–655.
Floridi, L. (2010). The Philosophy of Information. Oxford University Press, Oxford.
Ganascia, J.-G. (2007). Modeling ethical rules of lying with answer set programming.
Ethics and Information Technology, 9:39–47.
Ganascia, J.-G. (2008). In silico’ experiments : Towards a computerized epistemol-
ogy. Newsletter on Philosophy and Computers, American Philosophical Association
Newsletters, 7 (2):11–15.
Ganascia, J.-G. and Debru, C. (2007). Cybernard: A computational reconstruction
of claude bernards scientific discoveries. In Studies in Computational Intelligence
(SCI), pages 497–510. Springer-Verlag.
Gruber, T. (2004). Every ontology is a treat. AIS SIGSEMIS Bulletin, 1(3). Tom
Gruber interview.
Hatchuel, A. and Weil, B. (1995). Experts in Organizations: A Knowledge-Based Per-
spective on Organizational Change. Walter de Gruyter, Berlin-New-York.
Hutchins, E. (1995). Cognition in the Wild. MIT Press, Cambridge, Massachusetts.
Hutter, M. (2005). Universal Artificial Intelligence: sequential Decisions Based on
Algorithmic Probabilities. Springer.
Knoblock, C., Ambite, J., Carman, M., Michelson, M., Szekely, P., and Tuchinda,
R. (2008). Beyond the elves: Making intelligent agents intelligent. AI Magazine,
Manzotti, R. (2007). Towards artificial consciousness. American Philosophy Associa-
tion Newsletter on Philosophy and Computers, 07(1):12–15.
McCarthy, J., Minsky, M. L., Rochester, N., and Shannon, C. (1955). A Proposal for
the Dartmouth Summer Research Project on Artificial Intelligence. Technical report,
Dartmouth College.
Minsky, M. (1975). A framework for representing knowledge. In Winston, P. H., editor,
The Psychology of Computer Vision. McGraw-Hill, New York (U.S.A.).
Moravec, H. (1988). Mind Children. Harvard University Press, Cambridge, Mas-
Nelson, T. (1965). A file structure for the complex, the changing and indeterminate.
In Proceedings of Association for Computing Machinery 20th National Conference,
pages 84–100.
Newell, A. (1980). Physical symbol systems. Cognitive Science, 4:135–183.
Newell, A. (1982). The knowledge level. Artificial Intelligence Journal, 18:87–127.
Nobel, D. (2006). The Music of Life: Biology Beyond the Genome. Oxford University
Quillian, M. R. (1968). Semantic memory. In Minsky, M., editor, Semantic Information
Processing, pages 227–270. MIT Press, Cambridge, MA.
Ramalho, G., Rolland, P.-Y., and Ganascia, J.-G. (1999). An artificially intelligent
jazz performer. Journal of New Music Research.
Rickert, H. (1921). Kulturwissenschaft und Naturwissenschaft. J.C.B. Mohr (Paul
Siebeck), Tubingen, 5th edition.
Rolland, P.-Y. and Ganascia, J.-G. (1999). Musical pattern extraction and similarity
assessment. In Readings in Music and Artificial Intelligence. Contemporary Music
Studies, volume 20. Harwood Academic Publishers.
Rolland, P.-Y. and Ganascia, J.-G. (2002). Pattern detection and discovery: The case
of music data mining. In Pattern Detection and Discovery, pages 190–198. Springer-
Shapiro, D. and Goker, M. (2008). Advancing ai research and applications by learning
from what went wrong and why. AI Magazine, 29(2):9–76.
Simon, H. A. (1996). The Sciences of the Artificial. MIT Press, Cambridge, Mas-
sachusetts, 3rd edition.
Velcin, J. and Ganascia, J.-G. (2005). Stereotype extraction with default clustering.
In proceedings of the 19th International Joint Conference on Artificial Intelligence
(IJCAI), pages 883–888.
... Jika demikian, apakah AI menjadi bagian dari ilmu alam (natural science) atau ilmu sosialkebudayaan? Bagi Ganascia (2010), AI bisa menjadi bagian dari keduanya. Ilmu sosial mencoba untuk memahami tindakan-tindakan manusia, seperti tujuan manusia dan cara manusia mencapai tujuan-tujuan itu. ...
... Namun AI juga bisa menggunakan metode 'ilmu Jurnal Filsafat Indonesia, Vol 4 No 2 Tahun 2021 ISSN: E-ISSN 2620-7982, P-ISSN: 2620-7990 alam' misalnya dari matematika dan logika menyelidiki mekanisme fisiologis. Bagi Ganascia, seorang professor bidang computer science dari Sorbonne University, AI memang tampak sebagai bagian dari 'ilmu alam' [karena lahir dari generalisasi kasus-kasus melalui metode induktif], namun sekaligus juga bagian dari 'ilmu sosial' [karena mengkarakterisasi hal-hal artifisial berdasarkan fungsi, tujuan, dan adaptivitasnya, dan bukan hanya berdasarkan strukturnya melalui metode deduktif] (Ganascia, 2010). ...
... Kebudayaan mencakup wilayah yang luas karena juga menyertakan aktivitas interpretasi murni seperti sejarah. Lagipula, logika 'ilmu sosial-kebudayaan' merupakan perluasan logika yang digunakan dalam 'ilmu-ilmu artifisial' yang tetap memiliki kesamaan dengan 'ilmu-ilmu alam' (Ganascia, 2010). ...
Full-text available
Kecerdasan buatan (AI) adalah “payung istilah” yang digunakan untuk menyebut simulasi yang dilakukan oleh mesin-mesin atau alat, yang terhubung dengan samudera data, yang menyerupai kecerdasan manusia. Tidak diragukan lagi, AI sudah memberi dampak positif dalam banyak aspek kehidupan manusia: ekonomi, pendidikan, pemerintahan, hingga pertahanan dan keamanan. Namun, AI bagaikan dua sisi mata uang yang juga memberikan dampak negatif. Adanya dampak multidimensi yang ditimbulkan oleh AI membawa pada suatu pertanyaan tentang cara mengimbangi kemajuan AI agar tetap terarah pada koridor yang diinginkan. Untuk menjawab persoalan ini, filsafat selalu mengawalinya dengan analisis epistemologis. Dengan berangkat dari fakta dan analisis epistemologis tentang AI, penulis sampai pada keyakinan akan pentingnya peranan pendidikan interdisipliner. Dengan menggunakan metode kualitatif melalui analisis literatur tentang AI sebagai objek material, dan Filsafat khususnya epistemologi sebagai objek formal, tulisan ini menawarkan pentingnya AI dan ilmu Etika sebagai materi perkuliahan terpadu di tengah era dan disrupsi AI. Sebagai institusi pendidikan yang tanggap terhadap zaman, setiap universtas perlu mempertimbangkan perlunya materi perkuliahan tentang dasar-dasar AI dan Etika bagi setiap peserta didik.
... Out-of-context applications of science fiction ideas have also been criticized. For example, Jean-Gabriel Ganascia says that various technologies have been overhyped as a result of an abuse of the term technical singularity by Ray Kurzweil (Ganascia 2010(Ganascia , 2017. A humanities scholar, Jennifer Robertson, described the vision of the future depicted by the Japanese government as problematic, stating that it tended to confirm sexist representations inherited from the classic SF works (Robertson 2011). ...
Full-text available
Driven by the rapid development of artificial intelligence (AI) and anthropomorphic robotic systems, the various possibilities and risks of such technologies have become a topic of urgent discussion. Although science fiction (SF) works are often cited as references for visions of future developments, this framework of discourse may not be appropriate for serious discussions owing to technical inaccuracies resulting from its reliance on entertainment media. However, these science fiction works could help researchers understand how people might react to new AI and robotic systems. Hence, classifying depictions of artificial intelligence in science fiction may be expected to help researchers to communicate more clearly by identifying science fiction elements to which their works may be similar or dissimilar. In this study, we analyzed depictions of artificial intelligence in SF together with expert critics and writers. First, 115 AI systems described in SF were selected based on three criteria, including diversity of intelligence, social aspects, and extension of human intelligence. Nine elements representing their characteristics were analyzed using clustering and principal component analysis. The results suggest the prevalence of four distinctive categories, including human-like characters, intelligent machines, helpers such as vehicles and equipment, and infrastructure, which may be mapped to a two-dimensional space with axes representing intelligence and humanity. This research contributes to the public relations of AI and robotic technologies by analyzing shared imaginative visions of AI in society based on SF works.
... Our colleague Ganascia [28] defines the "spirit" of artificial intelligence as follows: "it shows how this epistemological view opens on the many contemporary applications of artificial intelligence that have already transformed-and will continue to transform-all our cultural activities and our world. ...
Full-text available
Can artificial intelligence (AI) be more ethical than human intelligence? Can it respect human values better than a human? This article examines some issues raised by the AI with respect to ethics. The utilitarian approach can be a solution, especially the one that uses agent-based theory. We have chosen two extreme cases: combat drones, vectors of death, and life supporting companion robots. The ethics of AI and unmanned aerial vehicles (UAV) must be studied on the basis of military ethics and human values when fighting. Despite the fact that they are not programmed to hurt humans or harm their dignity, companion robots can potentially endanger their social, moral as well as their physical integrity. An important ethical condition is that companion robots help the nursing staff to take better care of patients while not replacing them.
Full-text available
A growing debate in several European fora is paving the way for future rules for Artificial Intelligence (AI). A principles-based approach prevails, with various lists of principles drawn up in recent years. These lists, which are often built on human rights, are only a starting point for a future regulation. It is now necessary to move forward, turning abstract principles into a context-based response to the challenges of AI. This article therefore places the principles and operational rules of the current European and international human rights framework in the context of AI applications in two core, and little explored, areas of digital transformation: electronic democracy and digital justice. Several binding and non-binding legal instruments are available for each of these areas, but they were adopted in a pre-AI era, which affects their effectiveness in providing an adequate and specific response to the challenges of AI. Although the existing guiding principles remain valid, their application should therefore be reconsidered in the light of the social and technical changes induced by AI. To contribute to the ongoing debate on future AI regulation, this article outlines a contextualised application of the principles governing e-democracy and digital justice in view of current and future AI applications.
Full-text available
Our goal in this chapter is not to resolve or even attempt to analyze specific ethical issues that arise with AI. Rather, we will survey what we believe are the most important challenges for progress in the ethics of AI. At the present moment, there are many AI applications that are driving the interest in ethics; among them are autonomous vehicles, battlefield (lethal) robots, recommender systems in commerce and social media, and facial recognition software. In the near future we may have to grapple with disruptions in human social and sexual relationships caused by androids or with jurisprudence administered primarily by intelligent software. The developments in AI—now and in the foreseeable future—are sufficiently worrisome such that progress in the ethics of AI is in itself an ethical issue. The discussion of these challenges incorporates longstanding philosophical issues as well as issues related to computer science and computer engineering. We leave it to the reader to pursue technical details of both philosophical and scientific issues presented here, and we reference the background literature for such inquiries. The challenges fall into five major categories: conceptual ambiguities, the estimation of risks, implementing machine ethics, epistemic issues of scientific explanation and prediction, and oppositional versus systemic ethics approaches.
Computers and networks remind us how much our thought is orchestrated and they help us to understand that it always has been the case. In other words: the culture related to computers appears highly technical. Nevertheless, it is nothing more than the present translation of the set of know-hows related to literacy. This helps us discover (once more) an historical link between technical culture and culture of the scholars. Anthropologists have already explained how this culture of scholars and scientists extends to the culture at large: as a result of social domination (the power of writing) and by product of reflexivity: writing invites us to think about its objects and categories, and this meditation feeds culture in its wider sense. Hence, there is a direct link between technical culture related to writing and culture of a society.This helps us to show how digital and networked writing influences the problematics and epistemologies of disciplines commonly grouped under the label « human and social sciences »: new methods, combinatorial capacities, questions induced by the uses of the internet, and also basic skills (writing a text, finding a sign in a text). Some people claim that these questions are those of « digital humanities ». As scientists, we are more interested in the reasons as to why this movement is influential than in the weakness of their argumentation (fuzzy definition of DH, etc.): they reveal an interesting sociology of the University and manifest a loud silence of the representatives of social sciences about the way the frontiers of their disciplines evolve under the influence of present-day writing. However, the study of these changes, already studied by epistemologists, is very promising. And it helps understand the reality of digital culture, in its scholarly and popular expressions.
Full-text available
Personal motivation. The dream of creating artificial devices which reach or outperform human intelligence is an old one. It is also one of the two dreams of my youth, which have never let me go (the other is finding a physical theory of everything). What makes this challenge so interesting? A solution would have enormous implications on our society, and there are reasons to believe that the AI problem can be solved in my expected lifetime. So it’s worth sticking to it for a lifetime, even if it will take 30 years or so to reap the benefits. The AI problem. The science of Artificial Intelligence (AI) may be defined as the construction of intelligent systems and their analysis. A natural definition of a system is anything which has an input and an output stream. Intelligence is more complicated. It can have many faces like creativity, solving problems, pattern recognition, classification, learning, induction, deduction, building analogies, optimization, surviving in an environment, language processing, knowledge, and many more. A formal definition incorporating every aspect of intelligence, however, seems difficult. Most, if not all known facets of intelligence can be formulated as goal
In this article I argue that the best way to understand the information turn is in terms of a fourth revolution in the long process of reassessing humanity's fundamental nature and role in the universe. We are not immobile, at the centre of the universe (Copernicus); we are not unnaturally distinct and different from the rest of the animal world (Darwin); and we are far from being entirely transparent to ourselves (Freud). We are now slowly accepting the idea that we might be informational organisms among many agents (Turing), inforgs not so dramatically different from clever, engineered artefacts, but sharing with them a global environment that is ultimately made of information, the infosphere.
The 1956 Dartmouth summer research project on artificial intelligence was initiated by this August 31, 1955 proposal, authored by John McCarthy, Marvin Minsky, Nathaniel Rochester, and Claude Shannon. The original typescript consisted of 17 pages plus a title page. Copies of the typescript are housed in the archives at Dartmouth College and Stanford University. The first 5 papers state the proposal, and the remaining pages give qualifications and interests of the four who proposed the study. In the interest of brevity, this article reproduces only the proposal itself, along with the short autobiographical statements of the proposers.