=Paper=
{{Paper
|id=Vol-1315/paper5
|storemode=property
|title=How Artificial is Intelligence in AI? Arguments for a Non-Discriminatory Turing Test
|pdfUrl=https://ceur-ws.org/Vol-1315/paper5.pdf
|volume=Vol-1315
|dblpUrl=https://dblp.org/rec/conf/aic/Birner14
}}
==How Artificial is Intelligence in AI? Arguments for a Non-Discriminatory Turing Test==
https://ceur-ws.org/Vol-1315/paper5.pdf
How Artificial is Intelligence in AI?
Arguments for a Non-Discriminatory Turing test?
Jack Birner
University of Trento, University College Maastricht
jack.birner@unitn.it
Abstract. Friedrich von Hayek’s The Text Sensory Order (1952) presents a physicalist identity
theory of the human mind. In a reaction to Karl Popper’s criticism that such a “causal” theory of
the mind cannot explain the descriptive and critical-argumentative functions of language, Hayek
wrote a paper that was never published. It contains the description of a thought experiment of
two communicating automata. This paper confirms the impression of the AI-like character of the
structuralism and functionalism of Hayek’s Sensory Order. In some important respects, what
Hayek tries to do in his paper is similar to Turing’s discussion of the question “can machines
think?” Arguments will be given why according to a functionalist and physicalist identity theory
of mind the distinction between artificial and “natural” intelligence cannot be upheld. According
to such a theory, Turing tests are unnecessarily restrictive and discriminatory vis-à-vis
machines. In the end, the question whether or not machines can think is not meaningless, as
Turing thought. It can be replaced by the question if artificial minds are capable of
consciousness. The Turing test , however, cannot give the answer.
Key words. theory of mind ! physicalist identity theory ! virtual machines ! communication of
symbolic description
1 Introduction
This paper is the consequence of the interest in the philosophy of Karl Popper that I
share with Aaron Sloman. A couple of months ago he reacted to the announcement of
a conference on Popper that bears my signature and this led to both of us reading
some of the other’s publications. We discovered that we had more interests in
common. This happy chance meeting of minds led to my writing what you are now
reading.1 Popper is also one of the dramatis personae of this story, next to Friedrich
von Hayek. Popper and Hayek became close intellectual and personal friends during
and after the Second World War. In their published work they appear to agree on
almost everything. Some aspects of their thought, however, convinced me that this
could not be really true. And indeed, a closer look revealed that till the end of their
lives they remained divided on several important issues. I have dealt with some of
these, and with the influence – both positive and negative - they had on one another
elsewhere.2
1
Without Aaron’s encouragement I would not have dreamt of sending a text to a workshop on AI. Let me
hasten to add that his guilt stops here: I take full responsibility for everything that follows. I would also like
to apologize in advance for not referring to authors who may have discussed the same or similar problems;
these are my first steps in AI.
2
Cp. Birner (2009) and (forthcoming).
Page 61 of 171
2 Philosophy of mind
What I will take up here are their disagreements in the philosophy of mind. I do so
first of all because Hayek’s theory of mind and his defence against Popper’s criticism
have a strong AI flavour. 3 Second, there are some striking similarities between
Hayek’s work of the early 1950s and “Computing Machinery and Intelligence” (CMI)
of 1950 by Alan Turing, the third main character of this tale. These parallels deserve
more attention than has been given them.4 In 1952 Hayek published The sensory
order: an inquiry into the foundations of theoretical psychology (SO). The
foundations mentioned in the title are a philosophy of mind that I will now
summarize. Hayek tries to explain the human mind using only the laws of physics. He
had adopted this explanatory programme from Moritz Schlick’s Allgemeine
Erkenntnistheorie. The ontological idea underlying it is that the mind does not have a
separate existence from the brain. So Hayek’s is a physicalist identity theory.
As the vehicle for his explanation he uses a neural-network model.5 According to
Hayek, mental processes consist in the continuous reorganization on many levels of a
hierarchical system of relationships. That is why he speaks of an order of events. A
neural network is one possible model of the mind. Hayek is a radical functionalist in
the sense that he states that any physical configuration of elements and their
relationships might embody mental processes. He introduces this idea thus:
“That an order of events is something different from the properties of the
individual events, and that the same order of events can be formed from
elements of a very different individual character, can be illustrated from a
great number of different fields. The same pattern of movements may be
performed by a swarm of fireflies, a flock of birds, a number of toy balloons
or perhaps a flight of aeroplanes; the same machine, a bicycle or a cotton gin,
a lathe, a telephone exchange or an adding machine, can be constructed from
a large variety of materials and yet remains the same kind of machine within
which elements of different individual properties will perform the same
functions. So long as the elements, whatever other properties they may
possess, are capable of acting upon each other in the manner determining the
structure of the machine, their other properties are irrelevant for our
understanding of the machine.” (SO 2.28)
Then he proposes a radically functionalist and structuralist hypothesis:
“In the same sense the peculiar properties of the elementary neural events
which are the terms of the mental order have nothing to do with that order
3
Already hinted at in an afterthought to Birner 2009, where I wrote that Hayek missed the chance to be
recognized as a pioneer in AI. This will be discussed below.
4
But cp.Van den Hauwe (2011).
5
As does Donald Hebb, the publication of whose The Organization of Behavior in 1949 almost kept Hayek
from publishing his book. SO elaborates a manuscript that dates from 1920. For a discussion, cp. Birner
(2014).
Page 62 of 171
itself. What we have called physical properties of those events are those
properties which will appear if they are placed in a variety of experimental
relations to different other kinds of events. The mental properties are those
which they possess only as a part of the particular structure and which may
be largely independent of the former. It is at least conceivable that the
particular kind of order which we call mind might be built up from any one
of several kind of different elements – electrical, chemical, or what not; all
that is required is that by the simple relationship of being able to evoke each
other in a certain order they correspond to the structure we call mind.” (SO
2.29, my italics)6
This sounds very AI-like. The link between Hayek’s theory of mind and AI is even
more apparent in the way Hayek developed his ideas after the publication of SO. That
is the subject of the next section.
3 Popper’s criticism
Upon publication of SO Hayek sent a copy to Popper. Although Popper was – as
always - very polite in his reaction, he did not like it. Though Popper never writes this
down, his main general objection to Hayek’s theory of mind is that it is too
inductivist. What he does write in a letter to Hayek (2 December 1952) is that he
thinks his theory of the sensory order is deterministic. This implies, says Popper, that
it is a sketch for a deterministic theory of the mind. Now Popper had just written a
criticism (later published as Popper 1953) of this type of theory.7 He argues that a
deterministic theory of the mind cannot be true because it is impossible to have a
deterministic theory of human language.
In his criticism, Popper uses a particular analysis of language. He considers it to be
part of his solution to what he calls Compton’s problem. Popper uses that name for
what he considers to be a generalization of Descartes’ formulation of the mind-body
problem. Descartes asks how the immaterial mind can act upon the physical body.
Popper wants to know how abstract entities such as the contents of ideas and theories
can influence the physical world. He builds upon Karl Bühler’s theory of the
evolution of language. It says that the first function of language to emerge in human
evolution is the expression of subjective states of consciousness. The next function to
develop is communication (or signaling), followed by description. Popper adds a
fourth function, argumentation and criticism. It presupposes the previous (or, as
6
For a contemporary elaboration of this idea that seems to be very fruitful for understanding and measuring
consciousness, cf. Tononi 2012.
7
Apparently as a criticism of SO, of which he may have read the proofs. Cp. what Popper writes to Hayek
(letter of 30 November 1953 – Klagenfurt Popper archives, folder 541.12, on file from microfilm of the
Hoover Archives): “I was extremely pleased to hear that with “the challenge of my article on Language and
the Body Mind Problem”, I have done “a great service”. I am really happy about this article. I have ??? ????
M... (?) on the problem, but although I think that I got somewhere, I don’t know whether it is worth much.
If you really can refute my views (?), it would, I think, be an achievement.” (hand writing partially
illegible).
Page 63 of 171
Popper says, lower) functions. Not only has the need of humans to adapt to the
environment given rise to new physical instruments, it has also produced their
capacity to theorize. That is a consequence of the evolution of the higher functions of
language: they serve to control the lower ones (Popper 1972: 240-41). Abstract
contents of thought, meanings and the higher functions of language8 have co-evolved.
They help us control our environment “plastically” because they are adaptable.
Popper proposes a dualistic and indeterministic theory of the mind and of the
influence of the contents of consciousness on the world, which according to him can
account for the higher linguistic functions – unlike physicalist and behaviourist
theories:
“When the radical physicalist and the radical behaviourist turn to the analysis
of human language, they cannot get beyond the first two functions (see my
[1953]). The physicalist will try to give a physical explanation - a causal
explanation - of language phenomena. This is equivalent to interpreting
language as expressive of the state of the speaker, and therefore as having the
expressive function alone. The behaviourist, on the other hand, will concern
himself also with the social aspect of language - but this will be taken,
essentially, as the way in which speakers respond to one another’s “verbal
behavior.” This amounts to seeing language as expression and
communication.
But the consequences of this are disastrous. For if language is seen as merely
expression and communication, then one neglects all that is characteristic of
human language in contradistinction to animal language: its ability to make
true and false statements, and to produce valid and invalid arguments. This,
in its turn, has the consequence that the physicalist is prevented from
accounting from the difference between propaganda, verbal intimidation and
rational arguments.” (Popper and Eccles 1977: 58)9
Hayek took this criticism of Popper’s very seriously.10 He responded to it in “Within
Systems and about Systems; A Statement of Some Problems of a Theory of
Communication.” That paper was never published. It was never finished, either. Later
Hayek writes about it:
[I]n the first few years after I had finished the text of the book [SO], I made
an effort to complete its formulations of the theory in one respect. I had then
endeavoured to elaborate the crucial concept of “systems within systems”
but found it so excruciatingly difficult that in the end, I abandoned the
8
All of these are inhabitants of what Popper in his later philosophy has called world-3.
9
Popper & Eccles (1977) makes the same points that are made in the 1953 paper more forcefully.
10
“With the challenge of your article on “Language and the Body Mind Problem” you have unwittingly
done me a great service. Much has crystallized in my mind as a result of my inability fully to accept (?) the
argument. I believe I can now (?) provide (?) a causal theory of description and intention, but of course only
an “explanation of the principle” applicable to greatly simplified models and not sufficient to provide a full
explanation either of human language or human intention. But sufficient to construct models possessing all
the characteristics common to all instances of “description” and intention. I am still struggling with great
(?) difficulties, but I believe I am getting somewhere.” (Hayek to Popper, 30 October 1953, Popper Library,
Klagenfurt, folder 541.12, on file, from microfilm Hoover archives, hand writing partially illegible).
Page 64 of 171
longish but unfinished paper that apparently nobody I tried it upon could
understand”. (Hayek 1982: 290)
In the paper Hayek follows a two-pronged defence strategy against Popper’s
criticism, one “negative,” the other constructive or “positive”. As to the former,
Hayek states the purpose of the paper as
“deriving from the study of certain kinds of causal systems conclusions
concerning the character of our possible knowledge of mental processes. (…)
[T]he main conclusion to which [the argument] will lead is that for any
causal system there is a limit to the complexity of other systems for which
the former can provide an analogon of a description or explanation, and that
this limit necessarily excludes the possibility of a system ever describing or
explaining itself. This means that, if the human mind were a causal system,
we would necessarily experience in discussing it precisely those obstacles
and difficulties which we do encounter and which are often regarded as proof
that the human mind is not a causal system.” (Systems: 1).
Put bluntly, this “negative” part of Hayek’s reaction to Popper’s criticism is of the
heads-I-win-tails-you-lose type. The gut reaction of Popperian philosophers to such
an argument would be to condemn it out of hand as an immunizing stratagem.
Interestingly enough, Popper does not do so. I will briefly come back to this below.
The average non-Popperian citizen of Academe might instead dismiss it as corny.
That, however, would fail to do justice to Hayek. He gives two arguments for his
conclusion. First, as he states in the next sentence, “[w]e shall find that to such a
system the world must necessarily appear not as one but as two distinct realms which
cannot be fully “reduced” to each other.” (ibid.) The second argument invokes
complexity. In a generalized form it says that an explanans, in order to be successful,
has to be more complex than its explanandum. The argument is taken over from SO:
“any apparatus of classification must possess a higher degree of complexity than is
possessed by the objects which it classifies… therefore, … the human brain can never
fully explain its own operations.” (SO: 8.68).11 This may be true or false but it
certainly deserves closer examination. If it is true, then Hayek has demonstrated by a
reductio ad absurdum that the mind cannot explain12 itself (for it would have to be
more complex than it is).
The complexity Hayek refers to, and which he does not explain in more detail, may
consist of at least two circumstances. One has to do with problems of self-reference,
the other with the impossibility of describing all the relevant initial conditions for
explaining the human mind. Hayek does not mention or elaborate these aspects
(which would deserve closer scrutiny). What he does instead is to work out, in
subsequent publications, the methodological idea of in-principle explanations or
explanations of the principle, which are all we can achieve in the case of complex
11
For Hayek, who is a methodological instrumentalist, explanation is tantamount to classification. Cp.
Birner (forthcoming).
12
In the sense of classify, which is of course a view of explanation that is not shared by everyone (not by
Popper, for instance).
Page 65 of 171
phenomena. 13 Instead of rejecting this idea, that underlies Hayek’s “explanatory
impossibility theorem,” as part of a move to make Hayek’s naturalistic theory of mind
immune to criticism, Popper takes it seriously enough to refer to it 25 years later.14
In the modern literature on the mind-body problem Hayek’s argument is known as the
explanatory gap (cf. Levine 1983 and 1999 and Chalmers 1999). In SO Hayek claims
that his theory is less materialistic than dualistic theories because it does not assume
the existence of a separate mind-substance: ‘‘While our theory leads us to deny any
ultimate dualism of the forces governing the realms of the mind and that of the
physical world respectively, it forces us at the same time to recognize that for
practical purposes we shall always have to adopt a dualistic view’’ (SO, 8.46). This is
because we cannot produce a complete description or explanation of the processes
that constitute our mind and its relationships with the physical order without including
a description of the subset of those same processes that do the describing and
explaining, i.e., the mind itself. This again is because, as Hayek repeats in 8.44, his
theory is not a double-aspect theory. The complete order of all neural processes, ‘‘if
we knew it in full, would ... not be another aspect of what we know as mind but
would be mind itself.’’
Since SO is an identity theory, rather than denying the possibility of reducing the
sensory order to the physical order, it implies that there is no need to do so. In the
physical order, events are similar or different to the extent that they produce similar or
different external effects. In the sensory order, events are classified according to their
sensory properties: ‘‘to us mind must remain forever a realm of its own which we can
know only through directly experiencing it, but which we shall never be able fully to
explain or ‘reduce’ to something else’’ (SO 8.98). Yet, the two ways of describing
mental phenomena, in physical and in subjective terms, are two alternative ways of
describing the same phenomena. For the practical purpose of describing the mind
Hayek is a dualist in the sense that we humans with our human minds use different
languages describing the mental and the physical. Ontologically, there is just one
physical order.15
4 Hayek as a Pioneer of AI
13
Cp. for instance Hayek 1967.
14
“It has been suggested by F.A. von Hayek ([1952], p. 185) that it must be impossible for us ever to
explain the functioning of the human brain in any detail since “any apparatus … must possess a structure of
a higher degree of complexity that is possessed by the objects” which it is trying to explain.” (Popper and
Eccles 1977: 30).
15
Cp. Levine 1999: 11: “Metaphysically speaking, there is nothing to explain. That is, we are dealing with
a brute fact and there is no further source (beyond the fact itself) responsible for its obtaining. The fact that
we still find a request for an explanation intelligible in this case shows that we still conceive of the relata in
the identity claim as distinct properties, or, perhaps, the one thing as manifesting distinct properties. We
can’t seem to see the mental property as the same thing as its physical correlate. But though our inability to
see this is indeed puzzling, it doesn’t show, it can’t show, that in fact they aren’t the same thing. For what is
the case cannot be guaranteed by how we conceive of it.”
Page 66 of 171
The constructive defence against Popper’s criticism is undertaken in the second part
of the paper. Hayek describes a thought experiment that is meant to demonstrate that a
causal system is capable of one of the higher functions of language, description. By
“system” he intends
“a coherent structure of causally connected physical parts. The term system
will thus be used here roughly in the sense in which it is used in von
Bertalanffyi’s “General System Theory (…) [By system I intend] a persistent
structure of coherent material parts that are so connected that, although they
can alter their relations to each other and the system thereby can assume
various states, there will be a finite number of such states of which the
system is capable, that these states can be transformed into each other
through certain orderly sequences, and that the relations of the parts are
interdependent in the sense that if a certain number of them are fixed, the rest
is also determined.” (Systems, pp. 4-5)
Hayek concentrates on the behaviour of a type of causal system that he calls
“classifying system,” for a fuller explanation of which he refers to SO. 16 After
dealing, in the first part of the paper, with a series of preliminaries, Hayek is ready
with
“the setting up of the framework within which we wish to consider the main
problem to which this paper is devoted. In the next section we shall take up
the question how such a system can transmit to another similar system
information about the environment so that the second system will as a result
behave in some respects as if it had directly undergone those effects of the
environment which in fact have affected only the first system, but have
become the object of the “description” transmitted by that system to the
second.” (Systems: 18-9)
He introduces two automata 17 that communicate with one another by means of
symbols. Since he uses them in a thought experiment, it is justified to consider them
as virtual machines. 18 Hayek very ably concentrates on his main problem by
excluding the different problem whether, or to what extent, the structure of the two
systems have to be identical or similar in order to be able to interact with one
another:19 he assumes that they are identical. Hayek argues that the self-expressive or
symptom and signaling functions of communication pose no problem for his thought
experiment. Then he describes a situation in which the two systems are hunting a
prey. S1 can see the prey but S2 cannot because it is hidden from it by an obstacle. The
problem now is how S1 can describe and communicate the description of the itinerary
the prey is following to S2. The manuscript breaks off in the middle of this attempt to
fit the descriptive function of communication by means of symbols into the thought
16
Hayek’s description in SO of the human mind is that of a classifier system (a term he does not use).
17
Hayek does not use that term but he refers to Von Neumann’s theory of automata.
18
Aaron Sloman’s comment in correspondence.
19
He addresses that problem elsewhere. For a discussion, cp. Birner (2009).
Page 67 of 171
experiment, and in the framework of a causal theory of systems.20 Apparently he did
not succeed getting beyond the lowest two functions of communication.21 This is
precisely what Popper had said in his criticism.
5 Hayek and Turing
This section is dedicated to a (non exhaustive) comparison of the ideas in Hayek’s SO
and Systems with Turing’s in CMI. The objective is to give additional arguments that
Hayek’s SO and even more so his Systems deserve a place in the AI literature: if
Turing’s CMI is about AI, then so are these texts of Hayek’s.
5.1 What is the question?
In a comparison between Turing and Hayek we must not lose from sight that they
address different problems – at least at first sight. In CMI Turing poses the question
“Can machines think?” The problem Hayek wants to solve in SO is “What is
consciousness?” This, at any rate, is my reconstruction; Hayek himself is much less
sure and explicit in SO,22 even though he writes: “it is the existence of a phenomenal
world which is different from the physical world which constitutes the main problem”
(SO, 1.84). This is part of the qualia problem. It is different from the question whether
or not we humans can think; it is at best part of the latter problem. Nevertheless, the
way Turing and Hayek elaborate their respective problems show some similarities
that in my opinion make a comparison non futile.
Turing transforms his original question
“into [a] more accurate form of [it:] I believe that in about fifty years’ time it
will be possible to programme computers, with a storage capacity of about
109, to make them play the imitation game so well that an average
interrogator will not have more than 70 per cent, chance of making the right
identification after five minutes of questioning. The original question “Can
machines think?” I believe to be too meaningless to deserve discussion.”
(CMI: 442).
20
It breaks off in the middle of a word, “system”. That suggests that part of the typescript has gone
missing. I have repeatedly looked for the missing pages in the Hayek archives. A hand-written note by
Hayek on the first of the 27 typewritten pages of the ms. reads: “seems incomplete.” Added to Hayek’s
comment quoted in the third para. of section 3 above, this laconic note suggests that he has not looked very
hard for possible missing pages, which may be very few in number.
21
This is also suggested by the fact that years later Hayek writes to Popper that he feels “that some day you
ought to come to like even my psychology” (letter of 30 May 1960, Hayek Archives, Hoover Institution on
War, Revolution and Peace, box 44/2). This may be taken to imply that Hayek had not solved the problem
of showing that causal systems are capable of communication descriptions to other causal systems, thus
confirming Hayek’s comments (Hayek 1982: 290) quoted above.
22
This is highly uncharacteristic for Hayek, who in all his work follows a meticulously methodical
approach. Cp. Birner (2013).
Page 68 of 171
Now this reformulation comes much closer to the way in which Hayek elaborates the
problem of SO in the second part of Systems. His thought experiment, which is meant
to show that physical machines can express their internal states, signal, and
communicate descriptions to one another, qualifies as an early exercise in AI. That
exercise, moreover, is inspired by a physicalist identity theory of the human mind.
Turing’s “imitation game” is always interpreted as a procedure in which a human
mind attempts to debunk a computer that tries to imitate another human mind. A
generalized version of the game, one that is not based on the ontological assumption
that a human mind and a computer (and/or its software – in the sequel I will delete
this addition) are fundamentally different, would lose its purpose and become
meaningless. If there are no fundamental differences between computers and human
minds – as Hayek’s physicalist identity theory asserts – a Turing test would only
compare one kind of material realization of a mind with another. I will return to this
in the Conclusion.
When Turing discusses the possible objection of the “Argument from
Consciousness,” i.e., that machines can only be considered to be capable to think if
they are capable of experiencing feelings and emotions, he deals with the same
problem as Hayek in SO. Turing does not deny there is a problem, but he considers it
as different from, and secondary to, the problem that he addresses:
“I do not wish to give the impression that I think there is no mystery about
consciousness. There is, for instance, something of a paradox connected with
any attempt to localise it. But I do not think these mysteries necessarily need
to be solved before we can answer the question with which we are concerned
in this paper.” (CMI: 447).
Now, according to Hume “Reason is, and ought only to be the slave of the passions.”
(Hume 1739: 415).23 The very least we need for rational thought are motivations.24
Hayek deals with this effectively by describing how intentions may be modeled in his
thought experiment:
“By intention we shall mean such a state of a system that, whenever its
classifying apparatus represents a chain of actions as producing a result
which at the same time the internal state of the system singles out as
appropriate to that state, it will perform that chain of actions. And we shall
define the result or class of results which in any such state will activate the
chains of actions which will produce them as the goal or goals to which the
intention is directed.” (Systems: 17)
This is sufficient for the purpose of his thought experiment.
5.2 Functionalism
23
Research in cognitive science shows that Hume was right.
24
Aaron Sloman in correspondence.
Page 69 of 171
In the above, I have described Hayek’s functionalist approach to the mind. Compare
this with what Turing writes:
“The fact that Babbage's Analytical Engine was to be entirely mechanical
will help us to rid ourselves of a superstition. Importance is often attached to
the fact that modem digital computers are electrical, and that the nervous
system also is electrical. Since Babbage’s machine was not electrical, and
since all digital computers are in a sense equivalent, we see that this use of
electricity cannot be of theoretical importance. Of course electricity usually
comes in where fast signalling is concerned, so that it is not surprising that
we find it in both these connections. In the nervous system chemical
phenomena are at least as important as electrical. In certain computers the
storage system is mainly acoustic. The feature of using electricity is thus
seen to be only a very superficial similarity. If we wish to find such
similarities we should look rather for mathematical analogies of function.”
(CMI: 439)
This is identical to Hayek’s mental functionalism and structuralism.
5.3 Machines as subjects of themselves
When, on p. 449, Turing writes about machines being their own subjects, he seems to
have in mind a different problem than Hayek does when he addresses the question if
causal systems can describe themselves – by which he means fully describe.
“The claim that a machine cannot be the subject of its own thought can of
course only be answered if it can be shown that the machine has some
thought with some subject matter. Nevertheless, “the subject matter of a
machine's operations” does seem to mean something, at least to the people
who deal with it. If, for instance, the machine was trying to find a solution of
the equation x2-40a-11=0 one would be tempted to describe this equation as
part of the machine’s subject matter at that moment. In this sort of sense a
machine undoubtedly can be its own subject matter. It may be used to help in
making up its own programmes, or to predict the effect of alterations in its
own structure. By observing the results of its own behaviour it can modify its
own programmes so as to achieve some purpose more effectively. These are
possibilities of the near future, rather than Utopian dreams.” (CMI: 449).
This impression, however, may be mistaken. Compare the following passage:
“The idea of a learning machine may appear paradoxical to some readers.
How can the rules of operation of the machine change? They should describe
completely how the machine will react whatever its history might be,
whatever changes it might undergo. The rules are thus quite time-invariant.
This is quite true. The explanation of the paradox is that the rules which get
Page 70 of 171
changed in the learning process are of a rather less pretentious kind, claiming
only an ephemeral validity. The reader may draw a parallel with the
Constitution of the United States.” (CMI: 458)
This seems similar to the distinction Hayek makes, in para. 18, between changes
within a causal system and changes of the system itself:
“The concept of the state of a certain system must be carefully distinguished
from the changes in a collection of elements which turn it into a different
system. Different individual systems may be instances of the same kind of
system (or possess the same structure) if they are capable of assuming the
same states; and any one individual system remains in the same system only
so long as it remains capable of assuming any one of the same set of states,
but would become a different system in our sense. A full description of any
system would have to include sufficient information to derive from it
descriptions of all possible states of that system and of their relations to each
other, such as the order in which it can pass through the various states and
the conditions in which it will pass from one state into another. It will be
noted that strictly speaking a change in the permanent nature of one of our
systems such as would be produced by long term memory (the acquisition of
new connections or linkages) being an irreversible change implies a change
of the system rather than a mere change of the state of a given system.”
(Systems: 9-10)
The formulations are different, but Turing’s and Hayek’s ideas appear to be the same.
5.4 Hayek’s fate
Some of Hayek’s and Turing’s central ideas are very similar or even identical. Yet
Hayek has not been recognized as a pioneer of AI whereas Turing has. That might
have been different if he had published Systems. The radically thorough systematic
method that characterizes Hayek’s approach to each and every problem he ever put on
his research agenda25 kept him from doing so; he had, after all, failed to complete
what he considered to be the homework that Popper had assigned him with his
criticism of SO. Had he published the paper, even without a satisfactory account of
the communication of symbolic description between virtual machines, both Hayek
and AI might have been spared a lost opportunity.
6 Conclusion: for a scientifically and morally sounder Turing
test?
25
For an explication of this methodical approach cp. Birner 2013. This is not the only case of Hayek’s
being the victim of his own ambitiousness and thoroughness. Cp. Birner 1994.
Page 71 of 171
Perhaps the main defect of the Turing test as it is generally interpreted, is that it tests
whether humans have the subjective impression that machine intelligence is human.
As such, it may be of interest to psychology but hardly to AI. In addition, the Turing
test is biased or at least not general (and hence unduly discriminatory in the scientific
sense) in that it presupposes a particular type of theory of mind without making this
explicit, one that excludes the physicalist identity position. In CMI, C, the
interrogator, is a human being. In a scientifically sounder version of the Turing test
the population of humans and machines should be randomly divided in testers and
tested or judges and judged. But this would give rise to legitimate doubts as to what
the test is really testing. Is it the capacity of mind-like entities to recognize similar
mind-like entities?
There is no doubt that naturally evolved human minds and bodies are capable of much
more complex tasks than artificially created mind-like systems and their physical
implementations. This is not due to engineering problems in the realization of the
latter but to the fact that human minds and bodies are the products of a very long
evolutionary process. But we already know this without a Turing test.
Whether or not human judges in a Turing test can be fooled into thinking that
machine intelligence is human also depends on whether or not these judges think that
they share the same type of consciousness with the objects they judge. According to a
radical physicalist identity theory of mind machines are capable of having
consciousness and subjective feelings. If they don’t,26 this may be due to the fact that
we humans happen to have a longer evolutionary history, in which we have learnt to
have these impressions. Likewise, by interacting with humans, machines might learn
to understand and explain why we have subjective feelings (as in Star Trek). They
could even learn to have these impressions and sentiments themselves, particularly if
these have survival value (which in an environment that includes interaction with
human minds seems likely). The Turing test, however, is ill-suited for finding out
whether or not artificially created mind-like machines have consciousness, or have
consciousness that is similar to human minds. Giulio Tononi’s Integrated Information
Theory offers a much more sophisticated approach, one that even allows of measuring
the degree of consciousness – at least in principle. In this perspective it also seems
legitimate to ask if machines experience the same dualism as we humans do according
to Hayek (i.e. we cannot speak of the realm of the mental without using subjective-
psychological language;27 see above, the last two paragraphs of section 3).
The possibility that machines have consciousness may even raise an additional, moral,
objection to the traditional Turing test: it discriminates machines in favour of humans
26
But how could we find out? This raises the same problems Hayek addressed in Systems without finding a
solution.
27
The non-reducibility of a subjectivist language to a physicalist one that Hayek argues for may be seen as
a solution to what he considers to be a problem of complexity, viz. his explanatory impossibility theorem
(as I have called it). That is because subjectivist language enables us to speak meaningfully about mental
phenomena even in the absence of a complete reduction of them to an explanation in physical terms.
Perhaps the idea can be generalized to the question if subjective language and/or impressions may serve to
reduce complexity in general.
Page 72 of 171
by assigning the role of judges only to the latter. Machines might feel discriminated
against – if, I repeat, they are capable of moral feelings and other emotions at all.
So in the end, my arguments for a scientifically and morally sounder Turing test
seems to lead to the conclusion that the Turing test does not serve any useful purpose
at all. Turing’s belief, quoted above, that “[t]he original question “Can machines
think?” [is] too meaningless to deserve discussion” seems to me to be unfounded.
Thinking involves things such as intentionality, description, explanation,
understanding, creativity, having impressions and creativity. These are all features of
consciousness. So Turing’s question would reduce to the problem whether intelligent
machines are capable of consciousness. That certainly is a difficult question, but it is
hardly meaningless. As with so much research in AI, attempts to answer it have
taught us more about human minds than about artificial ones, and is likely to continue
to do so.
7 References
Birner J (forthcoming), “Generative mechanisms and decreasing abstraction”, in
Manzo (forthcoming)
Birner J (2014), F. A. Hayek’s The Sensory Order: An Evolutionary Perspective?
Biological Theory on-line first, DOI 10.1007/s13752-014-0189-4
Birner J (2013), “F.A. Hayek: the radical economist,”
http://econ.as.nyu.edu/docs/IO/28047/Birner.pdf
Birner J (2009), “From group selection to ecological niches. Popper’s rethinking of
evolution in the light of Hayek’s theory of culture”, in Parusnikova & Cohen 2009
Birner J (1994), “Introduction; Hayek's Grand Research Programme”, in Birner &
Van Zijp 1994
Birner J & van Zijp, R eds (1994), Hayek, Co-ordination and Evolution; His Legacy
in Philosophy, Politics, Economics, and the History of Ideas, Routledge, London
Bunge, M ed. (1964), The Critical Approach to Science and Philosophy. Essays in
Honor of Karl R. Popper, The Free Press, Glencoe
Chalmers, DJ (1999), “The Explanatory Gap. Introduction”, in Hameroff et.al.
(1999): 1-2
Hameroff SR, Kaszniak AW, Chalmers DJ eds (1999) Towards a science of
consciousness. The third Tucson discussions and debates. MIT Press, Cambridge
Hayek FA (1952) The Sensory Order. An Inquiry into the Foundations of Theoretical
Psychology. University of Chicago Press, Chicago, referred to as SO
Hayek FA (n.d.) “Within systems and about systems. A statement of some problems
of a theory of communication.” Hayek Archives folder 94/51, The Hoover Institution,
referred to as Systems
Hayek FA (1964), “The Theory of Complex Phenomena”, in Bunge (1964)
Hayek, FA 1967, Studies in Philosophy, Politics and Economics, University of
Chicago Press, Chicago
Hayek FA (1982) “The Sensory Order After 25 Years”, in Weimer and Palermo
(1982)
Page 73 of 171
Hume, D (1739), A Treatise of Human Nature, Selby-Bigge, LA ed., Claredon Press,
Oxford, 1896
Levine, J 1999, “Conceivability, Identity, and the Explanatory Gap”, in Chalmers et.al
1999: 3-13
Levine, J (1983), “Materialism and Qualia: The Explanatory Gap” Pacific
Philosophical Quarterly 64:354-61
Manzo G ed. (forthcoming), Paradoxes, Mechanisms, Consequences: Essays in
Honor of Mohamed Cherkaoui. Oxford: Bardwell Press
Parusnikova & Cohen RS eds. (2009), Rethinking Popper, Boston Studies in the
Philosophy of Science 272, Berlin, Springer
Popper KR (1953) “Language and the Body-Mind Problem”, in Proceedings of the
XIth Congress of Philosophy, 7: 101-7, Amsterdam, North Holland
Popper KR (1972) Objective Knowledge. An Evolutionary Approach, Oxford,
Clarendon Press
Tononi, G 2012, “Integrated information theory of consciousness: an updated
account” Archives Italiennes de Biologie, 150: 290-326
Turing, A (1950), “Computing Machinery and Intelligence,” Mind LIX (236): 433-
560, referred to as CMI
Van den Hauwe LMP (2011), “Hayek, Gödel, and the case for methodological
dualism,” Journal of Economic Methodology 18 (4): 387–407
Weimer WB and Palermo DS (1982) Cognition and the Symbolic Process, Hillsdale,
N.J, Lawrence Erlbaum
Page 74 of 171