RE: [agi] Google as a strong AI

[Ben Goertzel]

Paul,   I agree that there ARE paths to AGI that have the "valuable incremental deliverables" feature that you mention.  However, it seems to me that these are nowhere near the SHORTEST paths.   Simulating evolution seems to be a very long and arduous path to AGI.  It certainly took nature a hell of long time and a  hell of a lot of computing power.   The path we're following with Novamente will be  much quicker and less wasteful of resources -- if it succeeds.  But on the minus side, it possesses the "incremental deliverables" feature only weakly.   Fortunately, it does display the incremental deliverable features in a weak sense, i.e. many of the tools and components useful for Novamente are also useful for narrow-AI applications.   A better businessperson than I  could probably have already very significantly monetized a number of the existing Novamente tools/components, and with any luck even I'll manage to do it during the next couple  years.   -- Ben     >powerful applications -- e.g. that a 60%-complete AGI implementation should >be 60% as useful as a complete AGI.   It's just not true.  There are plenty of >other areas of science and engineering where this kind of "continuity" doesn't >hold either.  A 60%-complete spacecraft doesn't fly anywhere, and probably >has very little commercial value   -- so what?    Yes, this continuity certainly doesn't hold for space-crafts or a good many other engineering endeavors, but your analogy to them is flawed. If it were true, it would mean that it would be impossible for an AGI--more specifically, us--to evolve, since a 60% complete human-AGI, such as, say, a chimp, would presumably be too stupid to be successful in its environment, what with it only having a good deal less than 60% human-level AGI-power, after all. If evolution was able to make commercially successful spin-offs on its way to AGI, why can't you?   Paul Fidika [EMAIL PROTECTED]     To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/[EMAIL PROTECTED]