YKY (and Loosemore), You seem to be heavily influenced by cognitive linguistics theory. > Those people never come up with computational algorithms, all they do > is talk. You may have caught that disease too =)
Still, you like everyone else here is jumping into design before doing any serious analysis. It appears obvious (to me) that working a few problems through that in the range of both present people and future AGIs - some "desk checking" would point the way. I am no supporter of cognitive linguistics, and indeed I am quite in your corner here, but at least they do the analysis, rather than wasting time and money on designing and building non-functional systems. That they never get out of analysis simply shows that their approach doesn't work. However, people who never do the analysis are doomed to an even worse fate - designing and building systems with no hope of ever working. I still think the most promising approach is the logic-based one, but > I'll add special algorithms to take care of some of the phenomena > pointed out by cognitive linguistics. Right now I'm still learning > about it. Pick a real-world "problem" and see what YOU need to deal with it. Most interesting problems are NOT solved by some stroke of (AGI?) genius, but rather by embarking on a process that predictably leads to a solution. Underlying the countless "details" that I often challenge Ben and others on here, is a recognition that no one is doing any sort of decent job of analysis. Oops, I sense that Loosemore is about to jump in here and chew my ass, so in anticipation... *Richfield's Limitation of Cognitive Theory* A perfectly functioning system could be built any way at all that produced the perfectly correct behavior. When observing any system to determine its actual internal workings, you can only learn from the *difference* between perfect functionality and observed functionality. With really complex functionality, like that needed to compete in the real world, perfect functionality is often difficult/impossible to determine. The noise from the lack of knowledge of perfect functionality becomes greatly amplified when you then subtract off the observed behavior, so that most of what is left is noise, leaving little prospect of picking out many reliable observations from it *Corollary* Unless you are performing medical research to cure diseases, you are better off solving for perfect functionality, especially if you are trying to design AGIs that perform any better than people. Observations of people can certainly give you some good hints, but unless your objectives are medical, you should continue on to "solving" the real world problems. *Example Observation* ** When I was in high school, I cornered the football coach and gave him some simple 2x2 Game Theory problems set into football situations to solve by asking what he would do in these situations. His solutions were spot-on with Game Theory, as he explained how he would first decide the probabilities, and then select randomly among them. Clearly, simple Game Theory problems ARE something that at least some people are good at solving without any training at all, which of course tells us absolutely nothing at all about people, but it *does* tell us that real-world competition requires Game Theory skills. Steve Richfield ------------------------------------------- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244&id_secret=117534816-b15a34 Powered by Listbox: http://www.listbox.com
