starglider> Eric Baum wrote: >> However, the argument that it is not going to be possible, which I >> claim is at the least plausible, is (very briefly) the following: >> (1) understanding comes from Occam code, very concise code that >> solves a bunch of naturally presented problems (and likely, only >> from that).
starglider> While compact representations are important, I'm dubious starglider> about arguments that this is the only component of starglider> understanding. Understanding implies the ability to make starglider> predictions about configurations of a system that you starglider> haven't yet observed, the ability to generalise to other starglider> contexts and the ability to combine the model with other starglider> system models as part of a compound or hybrid starglider> model. Minimum-length descriptions aren't necessarily starglider> tractable for predictive purposes (certainly not optimal) starglider> or easily modified and combined with other models (I starglider> believe Ben Goertzel noted earlier that he makes a trade starglider> off in agent efficiency by using a common learning starglider> algorithm because the improvement in interoperability is starglider> worth it; presumably this includes starglider> larger-than-minimum-case representations and starglider> algorithms). Furthermore when it comes to probabilistic starglider> prediction and fuzzy compression it isn't immediately starglider> obvious what metric to use for preferring shorter, less starglider> accurate descriptions over longer but more accurate starglider> descriptions. The case for focusing on simplicity above starglider> all else is even weaker for solutions to non-predictive starglider> problems, though I'm not clear if your argument was that starglider> general. I believe these issues (as well as Richard's) are well addressed. What is Thought? has whole chapters about how exploiting structure is a distinct, hard problem from simply finding compact descriptions or representations. Occam code is posited to exploit structure, not merely be a compact description or representation. (Note in the above quoted sentence: the two word phrase "solving naturally presented problems". That's a compression of a chapter or two.) These matters are also briefly (but at greater length than is possible here) described in my paper A Working Model for General Intelligence at http://whatisthought.com/eric.html Unfortunately, I don't believe I have time or space to clarify in posts. BTW, in a previous post on this thread, I was unintenionally ambiguous, when I said What is Thought? was available in paper, I meant paperback. Sorry for the confusion. starglider> P.S. Interesting book, one of my favourite AI texts. starglider> Michael Wilson Director of Research and Development, starglider> Bitphase AI Ltd Web demos page: starglider> http://www.bitphase.com/apex ----- This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?list_id=303
