Something I used to ponder about teaching machines to be intelligent : there are always some 'bad' factors that an intelligent entity should avoid. For eg, an intelligent car should avoid hitting obstacles. So rather than seeing intelligence as actions arising it of the necessity to fit in, we can see it as actions that are necessary for survival. The AI in a tic tav toe should be able to learn how to play if it tried not to lose, that is, by trying to survive. As to what the cla can or cannot learn, there are more knowledgable individuals who can give you a better insight. From what I infer, the cla can learn anything - including basketball
-----Original Message----- From: "Chirag" <[email protected]> Sent: 14-02-2014 08:06 To: "[email protected]" <[email protected]> Subject: [nupic-discuss] I bring questions and AI product suggestions Hi Everyone, My name is Chirag. I joined this group recently and am actually quite excited about it. I think this is absolutely the right approach to developing AI. My primary profession is another field but I am pretty certain that AI will change the future. So I feel that by joining and contributing to this group is the best thing I can do for myself (fairly selfish view of life, I know). Given other obligations, I have limited time to contribute fully right now but I think I can contribute via my imagination and questions of what AI products could or should be created in the future. Below our some questions I have. Please let me know if I am thinking about the end game of true artificial intelligence correctly or am I selling AI short and there is more it will be able to do. Thinking part: When I immigrated to US as a youth, I had to learn a lot of different thing. I learned to speak English, pass school, play basketball and to buy appropriately styled clothes in the new culture(aka baggy pant etc). As a youth, I picked all these things up very quickly and my teacher said that I was smart to be able to pick up all cultural and academic knowledge in such a short-time. I think fear of not fitting in forced me to learn these things in a very short amount of time. But in any case, I was able to fool the world so to speak by imitating mostly. So for me one mark of intelligence is ability to learn anything/everything (via imitation/practice) out of fear of not fitting in. I am pretty certain AI products in future will contain these criterias of "What is not fitting in". I can delve into that later. Questioning part: So I think the question I have is Can the intelligent learning algorithms be used to learn anything? I think the simplest test can be can it learn to play Tic/Tac/Toe on its own (via trial/error/prediction)? and then can it play against me or a simulated computer? can it learn to play any given video game for that matter? Particularly, Can it learn how to play need for speed or any racing games? Product part: If our AI algorithms can learn to play video games (via trial/error/predictions), it can likely play the games in physical world too. Can the learning algorithm learn to drive a car (it's a game really, with fitting in criteria)? If it's figured out how to drive a car, can it pick me up at a given location at a given time? Can it learn to ride a bike? Can it learn to use a scateboard to get around? My favorite AI product of the future: A self learning robot which has figured how to play basketball. It can play against you or with you as part of a team (in real life). I am pretty certain this is the future we're headed towards at least of Artificially intelligent gaming. Best Rgds, Chirag -- Chirag
_______________________________________________ nupic mailing list [email protected] http://lists.numenta.org/mailman/listinfo/nupic_lists.numenta.org
