Blunder around for a while

Letters

Vol. 13 No. 24 · 19 December 1991

From John Skoyles

So Richard Rorty has joined Daniel Dennett (LRB, 21 November) on the connectionism-will-solve-all-our-philosophical-problems bandwagon. But like the similar enthusiasm for AI (Artificial Intelligence), this fervour is more a fashion created by what people think connectionism promises than a detailed understanding of what it can and cannot do. In the beginning AI promised much: but over time it has become apparent, as Rorty correctly observes, that it can only deliver what its programmers have already programmed into it. Rorty and Dennett enthuse over connectionism because it appears to be free of this problem. They are wrong. Connectionism suffers a similar, but undiscussed difficulty known as the training problem.

Essentially, connectionism concerns networks made up of interconnected units (and hidden units), each with adjustable ‘thresholds’ which learn generalisations by being trained upon exemplars and error correction. In somewhat simplified terms, networks learn by being given inputs which the network converts into outputs. Following the network’s output, the thresholds between its units are changed depending upon whether the output was correct or not. The training problem concerns where the network gets the information needed for this error-correction process. The problem is that error correction makes the training of a network dependent upon some kind of external assistance to tell it whether its output was the correct response to the input it was given: without this knowledge it cannot tell whether it has made an error or not and so cannot be trained. So while networks do not need programmers they do need some kind of external help.

Rorty and Dennett focus upon the wonderful generalisations networks can learn. They ignore where the external help needed to train networks might come from. But for many cognitive skills there is no easy way networks in the brain can obtain the error-correction feedback needed to train them. The problem is recursive: if there was a process in the brain which could provide this information then its own development would depend in turn upon some further process. For some cognitive processes like reading there may be ways out. Connectionist accounts of reading successfully account for human reading performance. It is likely that the error correction needed to train them comes in part from a separate kind of process which sounds out written words from their spelling (hence the difficulty in learning to read encountered by children with problems over sounding out words). However, there is no reason to assume consciousness is going to have such an easy way out of the training problem (and certainly Dennett has not provided one). It may be that even with connectionist models we will still find a need for God. Consciousness may have to remain unexplained for a bit longer.

John Skoyles
Psychology Department,

From Bob Grumman

If ‘what distinguishes a conscious state from a non-conscious state’ is, as Dennett claims, the former’s having ‘a higher-order accompanying thought that is about the state in question’, then what should we call the supposedly ‘non-conscious’ states experienced by animals aware of the presence of food, say, to distinguish those states from the non-conscious states that rocks and machinery and the like apparently experience throughout their existence? More important, what do we call that which experiences either a conscious or a non-conscious state of awareness if not a ‘consciousness’? And how can a Cartesian be wrong in supposing this ‘experiencing mechanism’ to be significantly different from its contents?

On a more mundane plane, I’d also like to know why Rorty disruptively goes against established usage concerning the generic third-person pronoun. Is it only to signal her sympathy for the neurotically hyper-offendable, or does she have some intelligent reason for it?

Bob Grumman
Port Charlotte, Florida

Vol. 14 No. 1 · 9 January 1992

From Jane Goodall

It’s a while since I’ve come across an example of that fine old genre, the Glowing Review that is also a narrative of mastery. But it’s made a major comeback in Richard Rorty’s piece on Dennett’s Consciousness Explained (LRB, 21 November). Who better than Rorty to stage it, with its all-star dramatis personae: Aristotle, Descartes, Hegel, Hume, Kant, Orwell, Stalin, Ryle – with Roger Penrose, John Searle and Thomas Nagel appropriately designated walk-on roles. And it’s stirring stuff. Dennett, student of a great master who ‘toppled settled philosophical convictions like ninepins’ in a work that ‘was rightly taken as the show-piece of post-war British philosophy’, has pulled off a remarkable feat. He has produced a work to equal it in boldness, originality and panache, full of ‘seminal’ arguments, which also display ‘an idiosyncratic mastery of metaphor’ etc, etc.

Well, this is all fine and for those who like that sort of thing it is most definitely the sort of thing they like. I don’t mind a good Helden-sage myself if I’m in the mood, as a matter of fact but that’s strictly provided they leave me – that is ‘she’ – out of it. I do object to being inscribed in the text when the roles for women are so lousy. What do I get here? ‘Aristotle, Ryle thought, had sensibly seen that to talk about someone’s mind is to talk about the features of her intellect and character …’ Penrose & Co agree that there are right answers to the question ‘ “What is immediately given to consciousness?”, where “given” is construed in such a way that the subject herself may be quite wrong about what is given to her consciousness’. ‘Even a reader who has never looked into a psychology book, or taken an interest in computers, will find herself absorbed in summaries of complicated psychological and physiological experiments, and in brief but clear accounts of curious computer programs.’ ‘Anyone sufficiently intrigued by this book to want a better understanding of this model will get what she wants from Andy Clark’s clear, helpful …’

After Aristotle has given my mind the once over and some of his distant descendants have surmised that I am wrong in the head, I get to be a B-stream reader who is to be offered brief, clear and helpful explanations about what goes on in the realms of seminal thought. Gee, thanks.

Jane Goodall
Bolton Point, New South Wales

From Peter Kalac

Richard Rorty is obviously of the opinion that Daniel Dennett has shouted ‘fire’ in a crowded Cartesian Theatre. If so, many will be expected to make a beeline to the nearest available exit. But on reflection (whatever that may be), Dennett is not convinced there is a Cartesian Theatre – no theatre, no exit – perhaps no fire either. In any event, I suspect John Searle and Thomas Nagel, regardless of where they are now seated, will indeed be hanging around for the second feature.

Peter Kalac
Red Bank, New Jersey

Vol. 14 No. 4 · 27 February 1992

From Leston Havens

Richard Rorty’s review of Dennett’s Consciousness Explained was lucid and helpful (LRB, 21 November 1991). I wish to refer those interested in its themes to Edward Hundert’s Philosophy, Psychiatry and Neuroscience, where similar ideas are developed.

Leston Havens
Harvard Medical School