Author: Graham Laight
Date: 02:09:11 06/28/01
Go up one level in this thread
On June 27, 2001 at 14:38:29, Dann Corbit wrote: >On June 27, 2001 at 07:26:10, Graham Laight wrote: > >>On June 27, 2001 at 05:59:05, Dann Corbit wrote: >> >>>On June 27, 2001 at 05:48:58, Adam Oellermann wrote: >>> >>>><snippety snip> >>>>>>You are correct Sir. A fly is more intelligent than the biggest computer. >>>>> >>>>>This is outdated information. Computer power in a single machine (multiple >>>>>processors) now exceeds the processing power of a fly. >>>> >>>>Yup, and if Moore's Law holds good for a few more decades, we might be able to >>>>manage a cockroach! However, the software implementation is bound to be buggy... >>> >>>Pretty amazing that a bee can do 20 GFlops on a few micrograms of nectar. Be >>>that as it may: >>> >>>http://cart.frc.ri.cmu.edu/users/hpm/project.archive/robot.papers/1999/SciAm.scan.html >> >>This article is way too pessimistic in assuming that human intelligence will >>take until 2050 to achieve. >> >>1. They assume that to do a task a human can do, you need the same amount of >>processing power. This is ridiculous. Chess computers are, or are nearly, at GM >>standard with only a tiny, tiny, tiny proportion of the processing power. A >>cheap chess computer, with a 10 Mhz processor, tiny amount of RAM, and 32k >>program will easily beat well over 99% of chess players. > >But it does not know it won and cannot tie a pair of shoelaces. > >>2. The article, at 2 years old, is massively out of date. Much of the article >>discusses the problem of a robot navigating its workspace. For a mere £35,000 a >>British hospital has just intalled a robot which has solved these problems - see >>http://www.ananova.com/news/story/sm_330824.html >> >>3. The article implies that the whole human brain is working as hard as the >>retina (the 2050 calculation is based on calculating how hard the retina works, >>weighing it, the multiplying by the weight of the brain). This is again >>ridiculous. >> >>You'll see - by 2025 there will be no aspect of human intelligence that can't be >>demonstrably outperformed by a machine. > >Personally, I don't think it is achievable. Human intelligence requires human >complexity. The programming of a system of that nature that is defect free >enough to function reliably is so far beyond the abilities of any team on earth >that it is ridiculous. OK then - show me a human whose "programming" is defect free. I don't think it's necessary to be "defect free" to be intelligent - I think you're mixing up AI with real-time control systems. >I think the horsepower will be there. I don't think there will be any way to >get all that power down to the road. Not only the horsepower - but also all the components. What will be required is to fit those components together into a single system image. If someone was to take a knowledge base approach to evaluating chess positions, they wouldn't need to build the whole thing from scratch - they could choose to make use of any of the plethora of ready made components from AI providers (or even information retrieval systems). Because the prevailing doctrine in chess programming is about speed, nobody has yet chosen this interesting approach, unfortunately. -g
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.