Path: utzoo!utgpu!news-server.csri.toronto.edu!rutgers!aramis.rutgers.edu!athos.rutgers.edu!nanotech From: autodesk!peb@uunet.uu.net (Paul Baclaski) Newsgroups: sci.nanotech Subject: Re: Some problems of super-intelligence Message-ID: Date: 7 Dec 90 08:51:46 GMT Sender: nanotech@athos.rutgers.edu Lines: 46 Approved: nanotech@aramis.rutgers.edu In article , dmocsny@minerva.che.uc.edu (Daniel Mocsny) writes: > ...Suppose a theoretical limit exists to the maximum amount of > intelligence that can exist in one coherent entity, before the > subparts become so intelligent that they create their own > independent agendas and rebel? Consider human organizations--the larger the organization, the more bureaucracy occurs. Minksky proposes in his Society of Mind (and in the epilogue of the new edition of Perceptrons (which I highly recommend for some critical analysis of connectionism)) that subparts would get gross overviews of what other subparts are up to. The more subparts and the higher the bandwidth, the more difficult this will be. The "subparts rebel" problem is probably closely related to the credit assignment problem in genetic algorithms. > [Actually, the million mark in increased intelligence probably is > the level we can expect to get *without* some fundamental increase > in knowledge about intelligence, simply by simulating the existing > structure but making it faster. Combine the raw speed with a > built-in library, and the resulting entity can apply to any problem > in 5 minutes the effect of 10 years of in-depth study and research, > current human scale. > --JoSH] This is a good point too. I can see many categories of "increased intelligence:" 1. Faster memory/procedures. 2. More concurency w.r.t. short term memory. 3. Higher sensory or output bandwidth. 4. More long term memory. All of these have been touched on recently. Additionally, there might be hard to quantify aspects like emotional responses like empathy and love or genius factors like insight and creativity. The distinction between long and short term memory is somewhat artificial since it does not include medium term activation. I suspect that the utility of such increases is dependent upon competition--a typical arms race. Paul E. Baclaski peb@autodesk.com