jsaxon@cs.tamu.edu (James B Saxon) (07/18/90)
Well, if nobody's going to do it, I will... This is a brief note about "The Emperor's New Mind" by Roger Penrose, Oxford University Press, 1989. He might know his physics but he does not have any clue about the brain or neural networks. He got Hebbian learning wrong! He said that the connection strength gets weaker with less activation (p. 397)! Now I was at the IJCNN '90 how many times did I hear 'pseudo-hebbian'? Ask Grossberg, ask Edelman. Ahem... And that's the least of it. Basically the arguments he poses in chapters 1, 9, and 10 have nothing to do with the rest of the book. But worse than that, he brings up points that have long been argued away... I'm going to leave these points unsubstantiated for now. If there's no argument, great. But, I wish to point out that we mustn't let creations like this go unanswered! This deserves a major rebuttal. It's up to us to not let the world continue thinking "Computers can only do what their programmers tell them to do." We must educate the people. Who is with me? -- ---- \ / ---- /--------------------------------------------\ James Bennett Saxon | O| | O| | "I aught to join the club and beat you | Visualization Laboratory | | | | | over the head with it." -- Groucho Marx | Texas A&M University ---- ---- <---------------------------------------------/ jsaxon@cssun.tamu.edu
cocteau@sun.udel.edu (Daniel J Pirone) (07/20/90)
Unfortunatly, I am of the school of thought that says, strengthen yourself, by knowing your enemy ( ie. I read perceptrons ... ;-)). So I bought, the Emporers New Mind, but like eveyone else in the Connectionist community, I have way to much to read.... So , ( on to the point ) I read the 1 page review in a recent (?) TIME ... I am not sure if it the journalism, or what, but he I got the idea that he was saying the neuron exhibit a quantum leap effect ??? ( an by some "GOD" given magic that only carbon based life forms with egos "understand", silicon objects can't .... ) This type of stuff is a double edge blade - nice fantacy reading, BUT REALLY BAD PR for our community.... Lynch him. thanks for you patience, the cybernetic cowpoke .
martin@oahu.cs.ucla.edu (david l. martin) (07/20/90)
In article <12788@sun.udel.edu> cocteau@sun.udel.edu (Daniel J Pirone) writes: >Unfortunatly, I am of the school of thought that says, strengthen yourself, >by knowing your enemy ( ie. I read perceptrons ... ;-)). >So I bought, the Emporers New Mind, but like eveyone else in >the Connectionist community, I have way to much to read.... >So , ( on to the point ) I read the 1 page review in a recent (?) TIME >... > >This type of stuff is a double edge blade - nice fantacy reading, >BUT REALLY BAD PR for our community.... >Lynch him. Oh come on, guys, what do you think this is, a football game, or a scientific dialog? I don't know anything about Penrose' personality, or what unspoken motives he might have. But I _have_ read his book, and I failed to detect any malice on his part towards the AI community. What I did detect, IMHO, was a thoughtful appraisal of some intriguing and important and timely questions, which by the way have occurred to a number of us already, who wished that we had the background in physics needed to answer them. Penrose has bothered to acquire the relevant background from a number of fields in addition to physics, in order to address these questions, and again IMHO, has done a very nice job of presenting the relevant concerns from each of these. Now, I agree with most that Penrose' conjectures don't seem all that plausible. Maybe this book ought to be considered the work of a brilliant man who's going through a starry-eyed phase. I don't know. But if one gives the book a fair reading, I think one will find that he's laid out the issues in a clear and scientific manner. Yes, there is a lot of speculation in the book. But where it occurs it is clearly identified as such. Moreover, the man has gone to great lengths to show how and why he believes these conjectures are related to the body of scientific knowledge. At the very least, he's performed a great service by defining the territory in which any further consideration of these questions can take place (or any refutation of his conjectures). Dave Martin UCLA
froncio@caip.rutgers.edu (Andy Froncioni) (07/26/90)
djp: Daniel J Pirone ( cocteau@sun.udel.edu ) --------------- djp: Unfortunatly, I am of the school of thought that says, strengthen yourself, djp: by knowing your enemy ( ie. I read perceptrons ... ;-)). djp: So I bought, the Emporers New Mind, but like eveyone else in djp: the Connectionist community, I have way to much to read.... djp: So , ( on to the point ) I read the 1 page review in a recent (?) TIME I think you definitely should make the effort to read the book rather than using a review from, of all places, TIME magazine... djp: I am not sure if it the journalism, or what, but he I got the idea that djp: he was saying the neuron exhibit a quantum leap effect ??? djp: ( an by some "GOD" given magic that only carbon based life forms djp: with egos "understand", silicon objects can't .... ) Again, I think you should read the book. The idea of self-similar systems possibly having an associated "generalised uncertainty principle" is not a totally new idea. In fact, this has been documented in literature concerning turbulence. The point is that there is always a portion of the phase space which is unknown to us, and in a sense, uncertain. djp: This type of stuff is a double edge blade - nice fantacy reading, djp: BUT REALLY BAD PR for our community.... If your "community" requires PR to sustain it, then you should really consider how much substance there is in it. Andy Andy Froncioni "It was a fine idea at the time, froncio@caip.rutgers.edu now it's a briiilliant mistake..." Elvis Costello
brucec@phoebus.phoebus.labs.tek.com (Bruce Cohen;;50-662;LP=A;) (07/31/90)
In article <6663@helios.TAMU.EDU> jsaxon@cs.tamu.edu (James B Saxon) writes: ... > > I'm going to leave these points unsubstantiated for now. If > there's no argument, great. But, I wish to point out that we > mustn't let creations like this go unanswered! This deserves > a major rebuttal. It's up to us to not let the world continue > thinking "Computers can only do what their programmers tell > them to do." We must educate the people. > > Who is with me? There has already been an excellent rebuttal in the form of an open letter from Hans Moravec, of the CMU Robotics Institute. I saw it on the USENET nanotech news group in February, and I believe it got fairly wide distribution at the time. It's fairly long, so I won't re-post it here. It's not necessary to make a jihad out of this debate; we may (and I do) think that Penrose' position is wrong-headed, but the best way to deal with this is to a) convince him of this with rational discourse, or b) prove him incorrect with our work. -- --------------------------------------------------------------------------- Bruce Cohen, Computer Research Lab email: brucec@tekcrl.labs.tek.com Tektronix Laboratories, Tektronix, Inc. phone: (503)627-5241 M/S 50-662, P.O. Box 500, Beaverton, OR 97077
pollack@dendrite.cis.ohio-state.edu (Jordan B Pollack) (07/31/90)
Andy Froncio writes: >Again, I think you should read the book. The idea of self-similar >systems possibly having an associated "generalised uncertainty >principle" is not a totally new idea. In fact, this has been >documented in literature concerning turbulence. The point is that >there is always a portion of the phase space which is unknown to us, >and in a sense, uncertain. I read it and found a warm and fuzzy (but unsellable) "Dancing Wu Li Masters/Tao of Physics" wrapped in a (sellable) Searle-like attack on the new "missing piece" of AI. (My capsule review: the Dancing Woolly in Surly Lion's Clothing - very transparent.) In Searle's case it is biological hardware dependency, and in Penrose's, it is quantum effects. Both miss the idea that identical functions can emerge from multiple forms. In AI, this is called the "software separability hypothesis", in Philosophy it is called "multiple realizability", and in evolutionary theory, it is called "convergence". For example, both flight and hierarchal social systems have arisen multiple times without any common ancestry. Why is consciousness always thought to be a "quantum leap" beyond other biological complex systems? Also, deterministic aperiodic systems are "uncertain" not because of a need for a quantum-level explanation, but because the amount of information about a system's historical state required to predict the next is too large for minds, mathematics, or computers. Finally, Nobel Prizes ARE the best PR, but only a few "communities" get them. If physics and chemistry don't require such PR, let them offer to relinquish their prize categories! One gets the feeling that this century's physics could come crumbling down when a lower level deterministic model is discovered for the uncertain phenomena in question. -- Jordan Pollack Assistant Professor CIS Dept/OSU Laboratory for AI Research 2036 Neil Ave Email: pollack@cis.ohio-state.edu Columbus, OH 43210 Fax/Phone: (614) 292-4890