"There would be an insidious problem with programming computers to play poker that in Sid's opinion would raise the Turing test to a higher level.
The problem would not be whether people could figure out if they were up against a computer. It would be whether the computer could figure out people, particularly the ever-changing social dynamics in a randomly selected group of people. Nobody at a poker table would care whether or not the computer would play poker like a person. In fact, people would welcome a computer, since computers would tend to play predictably. Computers would be, by definition, predictable, which would be the meaning of the word 'programmed. ' If you would play a computer simulation for a short amount of time, you would learn the machine's betting patterns, adjust would mean the computer would be distinguishable from a person. Many people would play poker as predictably as a computer. They would be welcomed at the table, too. If you would find a predictable poker opponent and would learn his or her patterns, you could exploit that knowledge for profit. Most people,however, have been unpredictable and human unpredictability would be an advantage at poker. To play poker successfully, computers would not only have to develop human unpredictability, hey would have to learn to adjust to human unpredictability as well. Computers would fail miserably at the problem of adjusting to ever changing social conditions that would result from human interactions. That would be why beating a computer at poker has been so easy. Of course, the same requirement, the ability to adjust unpredictability, would apply to poker playing humans who would want to be successful. You should go back and study how Sid had adjusted each hour in his poker session. However, as humans, we have been more accustomed to human unpredictability, so we have been far better at learning how to adjust." http://www.holdempokergame.poker.tj/adjust-your-play-to-conditions-1.html Of course, he's talking about dumb narrow AI purely-predicting-and-predictable computers, & we're all interested in building AGI computers that expect-unpredictability-and-can-react-unpredictably, right? (Wh. means being predicting-and-predictable some of the time too. The real world is complicated.). From: Jim Bromer Sent: Monday, June 28, 2010 6:35 PM To: agi Subject: Re: [agi] A Primary Distinction for an AGI On Mon, Jun 28, 2010 at 11:15 AM, Mike Tintner <[email protected]> wrote: Inanimate objects normally move *regularly,* in *patterned*/*pattern* ways, and *predictably.* Animate objects normally move *irregularly*, * in *patchy*/*patchwork* ways, and *unbleedingpredictably* . I think you made a major tactical error and just got caught acting the way you are constantly criticizing everyone else for acting. --(Busted)-- You might say my interest is: how do we get a contemporary computer problem to deal with situations in which a prevailing (or presumptuous) point of view should be reconsidered from different points of view, when the range of reasonable ways to look at a problem is not clear and the possibilities are too numerous for a contemporary computer to examine carefully in a reasonable amount of time. For example, we might try opposites, and in this case I wondered about the case where we might want to consider a 'supposedly inanimate object' that moves in an irregular and "unpredictable" way. Another example: Can unpredictable itself be considered predictable? To some extent the answer is, of course it can. The problem with using "opposites" is that it is an idealization of real world situations and where using alternative ways of looking at a problem may be useful. Can an object be both inanimate and animate (in the sense Mike used the term)? Could there be another class of things that was neither animate nor inanimate? Is animate versus animate really the best way to describe living versus non living? No? Given that the possibilities could quickly add up and given that they are not clearly defined, it presents a major problem of complexity to the would be designer of a true AGI program. The problem is that it is just not feasible to evaluate millions of variations of possibilities and then find the best candidates within a reasonable amount of time. And this problem does not just concern the problem of novel situations but those specific situations that are familiar but where there are quite a few details that are not initially understood. While this is -clearly- a human problem, it is a much more severe problem for contemporary AGI. Jim Bromer agi | Archives | Modify Your Subscription ------------------------------------------- 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=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
