|
Author
|
Topic: AI Perspective on the Cell (Thoughts on Jonker et al. 2002)
|
James A. Barham
Member
Member # 50
|
posted 12. April 2002 10:57
I have just finished reading the important new article, "Putting Intentions into Cell Biochemistry: An Aritificial Intelligence Perspective," by Catholijn M. Jonker and her co-workers (J. Theor. Biol., 2002, 214: 105--134), and since this is an article that should be of interest to everyone here at ISCID, I wanted to help publicize it, and also share the reflections to which it has given rise.
The paper may be accessed at : http://www.cs.vu.nl/~jonker/cve/publ.html#Journals (under "2002, item 2")
In this paper, Jonker et al. establish beyond the shadow of a doubt the practical utility of applying what they call the "Belief, Desire, Intention" (BDI) logic to the functioning of the living cell. This idea in itself is far from new. Howard Pattee, for one, has been arguing that this is necessary (and irreducible to chemistry) for something like the past forty years. What is new is the rigor and the depth with which Jonker et al. establish that the BDI methodology not only works, but is in fact indispensable for understanding the way in which individual chemical reactions are organized in cellular metabolism.
My reflection is this. We have just spent the better part of fifty years getting out from under the fallacious and question-begging doctrine of representationalism in AI itself, in cognitive psychology, and in the philosophy of mind. People like Hubert Dreyfus, John Searle, Walter J. Freeman, Timothy van Gelder, Alicia Juarrero, Fred Keijzer, Sunny Auyang, and many others have successfully shown that representationalism is a mistake, and that the brain must be approached, not as a symbol processing system, but rather as a dynamical system. This battle is now all but won (Hilary Putnam's defection was the first sign that representationalism was in serious trouble, but the official capitulation was Jerry Fodor's mea culpa in his most recent book).
Now, if it is true, as Jonker et al. say, and as I also believe, that the cell exhibits intelligence in precisely the same sense as the brain does, then shouldn't the lessons we have learned from the history of AI be applicable to the cell? By ignoring this history, aren't Jonker et al. in danger of repeating it? They speak unabashedly in the old language of representations and symbols, which brain researchers themselves are now abandoning in droves. There has to be a better way.
But what is it? I don't pretend to know the answer to that question, but I would like to share one more thought.
Yesterday, I also chanced upon the outline of the new book by Robert Laughlin and his ICAM colleagues, by following the link on Metanexus.
Here it is: http://www-ccrma.stanford.edu/~lew/
This book, which will clearly be a landmark achievement when it comes out, contains an interesting (if frustratingly elliptical) discussion of the emergence of "Higher Order Principles" throughout physics, and so, one must believe, in living systems, as well. I submit that Laughlin et al. are presenting us with the key necessary to interpret Jonker et al. correctly, and to help us avoid committing the mistakes of AI all over again at the cellular level.
Boy, both these pathbreaking works in a single day. . . We are truly living in amazing times.
[ 13 April 2002, 08:34: Message edited by: James A. Barham ]
IP: Logged
|
|
John Bracht
Member
Member # 5
|
posted 12. April 2002 12:46
James,
I'm familiar with the Jonker et al. article, having read it a few months ago (we even cited it in the ISCID news section). But not being an A.I. expert, and having only the slightest introduction to the issues you mention, I would appreciate a clarification of what, exactly, representalionalism is and how it is flawed in the case of A.I., and thus how it may be flawed in its application to the cell. It would also be interesting to hear your take on how we are to avoid the representationalism mistake when "putting intentions back into the cell." What are your thoughts on this?
Thanks,
John Bracht
IP: Logged
|
|
James A. Barham
Member
Member # 50
|
posted 13. April 2002 08:23
John:
There are two issues here, the question of the need for symbols (or representations---I am treating the two terms as synonymous) to understand intelligence in general, and the question of how to explain the natural ground of the teleology (or normativity) that we observe in living things. For this post, let me limit myself to the first question (I have dealt with the second one at length already elsewhere on Brainstorms).
First of all, it is clear that human beings do use symbols when we speak (and read and write---but pace Derrida, I think we can ignore the late accomplishments of civilization in this context). But the question is, Does it make sense to explain the operations in the brain itself in terms of symbols and representations?
Well, What is my definition of a symbol? Just the standard Peircean one: An entity that stands in a triadic relation, mediating between an object represented (signifie) and an "interpretant" (in Peirce's vocabulary), that is, a cognitive agent who interprets the symbol (signifiant) as representing the object (signifie). The key point is that there are no symbols in the absence of cognitive agents to interpret them. Another key point is that there is no intrinsic relation between the symbol and its material instantiation. That is, there is no intrinsic relation between the acoustic pattern of the English word "table" (or the Modern Greek word "trapezi", etc., etc.) and the thing itself. Another example I like to use is the relation between a green light and the rules of traffic, or in shorthand form, "Green means go."
Now, it was pointed out a long time ago (by Searle, Harnad, and others) that it is not possible to explain human cognition using the idea of symbols, because either it creates an infinite regress---that is, it requires another mind at a deeper level (the famous "homunculus")---or else it implies dualism, i.e., the interpretant standing outside the brain altogether (immortal soul, God, etc.). At any rate, if the brain operates with symbols, then you need another interpretant to interpet the physical tokens in the brain AS symbols.
But you may say, Wait a minute, Barham, aren't you agreeing that the cell is intelligent? Therefore, don't the processes within the cell have purpose and meaning? And if they do, how else can we understand them, if not as symbols?
Certainly, the dominant viewpoint in the philosophy of mind is that all functional processes are symbolic in nature and therefore lack the intrinsic connection between meaning and physical instantiation (i.e., they are "multiply realizable", in the jargon). But this is not necessarily true. Take the case of "sweet means eat." Here, the very same molecule that is in fact good for us metabolically, also tastes good. There IS an intrinsic link between the meaning of the sweet taste and the physical basis of the taste. Now, there may well be cases in the cell ("second messengers" and so on) that are more like "geen means go," but if so, they can only be understood as being instrumental to other processes that have intrinsic value (as in "sweet means eat"). Somewhere, there has to be intrinsic value. In short, the philosophical doctrine of functionalism is fundamentally misguided.
All of this is pretty well accepted by now at the level of the brain (see, especially, Timothy van Gelder, "What Might Cognition Be, If Not Computation?", Journal of Philosophy, 1995, 91: 345--381). All I am saying is that let's not commit the same mistake all over again at the level of the cell by introducing symbols where none are needed. The only true symbols that exist in the world are at the level of human language. At infra-linguistic levels, we must assume that there is a different kind of non-symbolic teleological control going on, where the connections between meaning and matter are direct and unmediated (non-representational)---although filling that out in a scientifically respectable way is of course another matter.
[ 13 April 2002, 08:35: Message edited by: James A. Barham ]
IP: Logged
|
|
|
Printer-friendly view of this topic