On Fri, Jul 14, 2006 at 12:07:55PM -0400, Eric Baum wrote: > > Warren Smith recently pointed out that if you regard a CMOS transistor > pair as roughly comparable to a synapse, and assume rather generously
Groan. The whole network computes. The synapse is just an element. Also: you're missing on connectivity, reconfigurability, synapse type and strength issues. > that synapses can continuously operate at 400 Hz, a 3.6 GHz Xeon with > 286E6 transistors has processing power .5 X 3.6 X 286 X E(9+6)=5.1 E17 In principle, a wafer full of ring oscillators each at THz is best at converting Watts into meangingful OPS. If we maximize the bit change rate, the thing obviously computes, no? > whereas a human brain with between 1E12 and 1E15 synapses > has "clearly inferior raw power" 4E14 - 4E17 bit-ops/second. If I have a single bit flipping ~THz rate, is this the same amount of computation as a 10^12 bits computing a meaningful computation, each operating at ~Hz? Obviously not. > "Plus the Xeon bit-ops are better understood and probably more > reliable than the human bit-ops. These bit-op/sec estimates are of > course merely upper bounds on what is achievable in practice", because > your brain would fry if all synapses fired at 400hz simultaneously, > and likewise the Xeon. Er. My brain has just bein fried, by reading above passage. > If somebody out there has some strong reason why the above is > misguided, I'd be interested in hearing it. > > On a related subject, I argued in What is Thought? that the hard > problem was not processor speed for running the AI, but coding the Trust me, the speed is. Your biggest problem is memory bandwidth, actually. > software, and estimated that evolution had gone through very roughly > 10^35 viruses since the dawn of time, as very rough estimate of raw power > available to it for software discovery. You certainly need a lot of ops to design an evolvable system. Most people are still unaware that this is the critical threshold in making AI by evolutionary algorithms. > Dan Fischer did a similar physicist's estimate, coming up with 10^35 > bacteria. > However, Warren has recently done some digging > on the subject, and come up with what seems to be a better estimate > that 10^44 bacteria have lived on Earth. -- Eugen* Leitl <a href="http://leitl.org";>leitl</a> http://leitl.org ______________________________________________________________ ICBM: 48.07100, 11.36820 http://www.ativel.com 8B29F6BE: 099D 78BA 2FD3 B014 B08A 7779 75B0 2443 8B29 F6BE ------- To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/[EMAIL PROTECTED]
signature.asc
Description: Digital signature
