[Serious Phil] Presupposing Experiential Subsystems
larry_tapper_2 at yahoo.com
Thu May 31 13:16:35 CDT 2012
PDJ > > In my last reply I omitted to note an interesting omission. In MB&P, Searle notes three Replies to the CRA. None of them is direectly a complaint that the CR is underspecified. Why would no one have gien him that response if it is so obvious?
PDJ> How about answering the question, Stu? "Why would no one have given him that response if it is so obvious?"
SWM> "Because philosophy is very complex and subtle and even smart people take a while sorting the issues all out at times. There is a very strong intuition on Searle's side. However, I agree with Dennett that the system reply is the right one but the early versions of it that I've seen missed the point that the CR qua system fails on the basis of the specs Searle gives it.
IIRC you have the facts quite wrong here.
As I recall, what you call the "underspecking" response was a quite commonplace response to Searle's original statement of the CRA, from day one. I am quite sure, for example, that we can find Hofstadter making this point back in the 70s when I was following these things more closely. Hofstadter pointed out that a system that could pass the Turing Test would have to be far more complex than the mere squiggle-matcher Searle apparently described in his early presentation. This is not a terribly subtle point, of the sort that would take decades to notice.
PB's much vaunted (by you) Bicycle Reply says absolutely nothing that Hofstadter didn't say 35 years ago.
So why didn't Searle list underspecking as one of the major early objections to the CRA? The simple answer IMO is that it *wasn't* a major objection to the CRA. It was just a most likely justified complaint about a misleading aspect of Searle's presentation. As I argued in my previous post, if the original CRA was underspecked, that is only, at worst, because Searle attempted to pull a rhetorical fast one, exaggerating the simplicity of AI software designs to sway a wider audience.
But we all know perfectly well that if we add complexity to the CR software as described, Searle would (and did) still make essentially the same argument, based on the intrinsic nature of software. In that case, why make such a fuss about underspecking? The only point of it is apparently to chastise Searle for rhetorical trickery in his original presentation --- the real issues are not being touched.
I am not taking sides here in the matter Dennett v. Searle. I agree with you, of course, that how we define 'consciousness' is part of the crux of the matter, as well as how it could be an emergent property of a system with insensate parts. All I am saying is that the underspecking argument, as prezented by you, is rather pointless, and that the problem has been compounded by your dodgy presentation of the "upspecking" that would need to be done.
Your pal Bill Modlin understood all this perfectly, and in my opinion he completely nailed it in the Analytic post, addressed to you, which I quoted earlier:
"What I pointed out was that Searle has explained, and apparently has
intended from the beginning, that his CR "rules" are to include any sort of complexity that anyone wishes to add. We do not need a "more complex system" as he says to assume it is already there, that we are free to use any specification of our choice.
"He says that no system of "mere computations", no matter how complex, will ever come to understand what it is doing. He feels that computation is the wrong sort of thing to produce understanding, that there is "something missing" that cannot be supplied by any elaboration of the computational scheme.
"Your arguments have been mostly about adding different more powerful
computations to what you interpret as his intended implementation, but since his expressed intent was to allow all that you say is needed, and his assertion is that this does not change the argument, then you are not addressing his argument.
"Worse, you still do not seem to have fully appreciated that most of
your suggested "improvements" would not make a real difference. You confuse hardware with algorithms, and still seem to feel that somehow there are things that are possible with a complex parallel physical mechanism that are not possible with a simple sequential one. Which is false, and makes most of your arguments pointless.
"I think Searle is wrong, for reasons that have nothing to do with complexity or speed....
[various criticisms of Searle follow]
"...If anyone cared we could probably elaborate more ways that he is wrong, but I have become tired of this discussion and don't feel like bothering."
Moreover, Dennett, in the early nineties in his book Consciousness Explained, goes to the heart of the
> matter without explicitly speaking of underspecking when he writes that the reason the CR fails is because it is insufficiently complex. So, in fact, the early years of this debate (which began with Searle's formulation of his CRA in the eighties) did not miss the point I'm making. What seems 'so obvious' now was less obvious before all the confusions built into the argument had been sorted
> Philscimind mailing list
> Philscimind at ...
More information about the Philscimind