From ccbanzai@showme.missouri.edu Tue Feb 10 17:49:00 1998 Date: Mon, 26 Jan 1998 12:21:46 -0600 (CST) From: Jim Heivilin Reply-To: blue_planet@mpgn.com To: blue_planet@Phaser.ShowCase.MPGN.COM Subject: Re: [BLUE PLANET] - AI Date: Mon, 19 Jan 1998 10:04:36 +0100 From: "Cedric M. BoB PONCHY" To: blue_planet@MPGN.COM Subject: Re: [BLUE PLANET] - AI I've been following this thread for a while. Time to add a few comments. >Date: Mon, 19 Jan 1998 08:10:33 +0200 >From: dreamwvr > >Firstly, I'd like to suggest Isaac Asimov's robotic shorts for those who >are interested in how robots or AI can screw things up. Good books. Let's not forget they've been written a LONG time ago ;) >Second on AI, I think the main focus is EXPERT SYSTEM and not intelligence. >It is programmed to handle as many situations as possible, and possibly has >limited learning ability, but it has definite limitations. Definitely true. It took them more than 20 years to realize that AI is a dead-end. Despite the big bucks and the support of powerful organizations, AI never reached it's objectives. AI pretended to be a new fundamental science, like Maths or Physics. AI gurus were very disdainful of other computer sciences and rejected without thinking new ideas. They utterly failed to fulfill their goals. On the bright side, AI researchs gave birth to many useful tools (neural networks, expert systems, etc...) that are the corner stones of modern computer sciences (and are really useful in Real Life). >Decision processes you and I take for granted, are horrific for a computer >to handle. >ie. If we were to consider that case of a computerised taxi cab. Things >that we take for granted that would need to be programmed are: >- Get the passengers there safely. >- Stay on the road. >- Do not use the pavement. >- drive within speed limits (although humans seem to have this problem too) >- Don't break so hard that the passengers end up on the front window. >- Unexpected situations? What to do. >etc. . . The problem when you write a system expert is that you have to list EVERY elementary task. In the best case its really hard. In most cases, it's neraly impossible, because a lot of subconscious tasks are involved. The human walk is impossible to simulate with classic techniques without really big computers. In the late 50's, when AI began to emerge, some people set in motion a chain of events that finally gave birth in the late 80's IIRC to Artificial Life. AL takes the problem by the other end: it studies the evolution of artificial creatures in simple (for the moment ;) ) universes with a few fundamental rules. It gave tremendous results, and when biologists joined the party it really began to kick butts. The idea is that it's easier to built a virtual universe, with a set of fundamental rules, to put some creatures in it with a goal and see how they're evolving to reach the objective, than try to find every mechanism or function to perform the task. The Conway Game (Life game?) is such a virtual universe. With some sweat, it's possible to use this cellular automaton to build a complete computer (von Neuman IIRC demonstrated it). Some weirdos even designed artificial universes to solve some programming problems. The function's code was the creature, and they found valid solutions that were really different from code designed by human programmers. It's a cheap and efficient way to solve complex problems. So, the idea is that the more powerful the computer, the more complex the simulation; so I think the AI is not far away from becoming an everyday reality. To get back to the taxi example, it could be achieved relatively easily: put the traffic regulation plus the necessary physics together as the rule of the universe, and run the sim; maybe after a few million cycles, you'll get something that will safely deliver the passenger at the right place :) An important point is the distinction between Intelligence and Conscience. The level of Conscience is what really makes the difference between man and animals. Animals are intelligent; they can learn to perform complex tasks, they can calculate, etc... I believe we're actually able to design artificial creatures able to perform a lot of non-trivial task with a terrific efficiency: keep the house warm, regulate traffic, etc... But conscience is, for me quite different: it's a sort of filter; it filters our perceptions and determine how many informations we get from our perceptions so the higher the conscience, the more the informations, got the idea ? Well, it's still very fuzzy for me yet :) Conscience would require not only perceptions to register the experiences, but also means to manipulate and experiment the environment. It would also need a support with a very high level of complexity. To make it quick, you need a big brain (ie. a very complex perception/appendages processing unit) to attain a high level a conscience, but a big brain doesn't automatically gives you Intelligzence and Conscience. I'd recommend the book "Destination: Void" of Franck Herbert; I think there is a very valid point in this book that is confirmed by recent advances in Artificial Life researches. Lastly, as somebody else said on the list, a complex AI would be truly "alien" to us, and maybe we couldn't recognize it as "intelligent". So what in the BP universe ? I'd say that AI exist; they're as common as transistors nowadays. Every automated task requiring some adaptation would use some kind of AI, actually some artificial creature :) because it would give better results and should be very cost effective to develop. Human sized Intelligent&Conscient androids? Probably no. First for technical reason: does the tech level of the BP universe allow the level of integration necessary for this? Next, for more mundane reasons: what would such a creature think of humans? what would be its attitude (indifference, hostility...)? I think that the odds of getting a somewhat human profile are very slim. Should it happen, it would be pure luck IMHO. So, such a creature would pose a real problem to our society. Huge black ops giant sized computer experimental AI/Conscience? Possible. But the question is still there: WHAT would it be, what would be it's attitude toward humans. It could give some very interresting plots: a powerful Corpo has such a creature in a secret lab somewhere, and it get rid of its masters and pursues its oan agenda... Got the idea? Any comments ? -- Il y a deux façons de faire les choses: La mauvaise La mienne -- Cédric PONCHY, The BoB Master ///\Oo/\\\ ponchy@club-internet.fr *************************************************************************** To unsubscribe from this list send mail to majordomo@mpgn.com with the line 'unsubscribe blue_planet' as the body of the message. From BIOHZD@aol.com Tue Feb 10 17:49:00 1998 Date: Mon, 26 Jan 1998 20:26:12 -0500 (EST) From: BIOHZD@aol.com Reply-To: blue_planet@MPGN.COM To: blue_planet@MPGN.COM Subject: Re: [BLUE PLANET] - AI Hello Cedric, Your characterization of AI in BP is very close to what we are working on. As I said a few posts ago, check out our upcoming book, Archipelago. The city of Dyfedd described therein will give those of you interested in AI a place to try out your ideas. Jeff Barber Biohazard Games *************************************************************************** To unsubscribe from this list send mail to majordomo@mpgn.com with the line 'unsubscribe blue_planet' as the body of the message.