If someone made a machine that, for example, plays poker, could it learn from its mistakes in one game and not make them in the next? Or could it learn certain moves in chess from someone using them against it?Can machines and computer learn?
Yes, they can learn up to a point. The spam filtering in most email programs use learning, for example, to recognise new kinds of spam that they've never seen before.
Also, some anti-virus programs use a heuristic, which is kind of like an advanced guess, to find new viruses. Combined with sending the found virus back to the virus company to be confirmed and included in new versions of the program, that's a KIND of learning. Not the best example, but something that happens every day.
Some school timetabling programs use learning -- they actually might evolve the answer using a simulation of creatures. Each "answer" (each organisation of times/classes/teachers/students) is considered to be an animal. They then "breed" all these different animals, and the ones that are best (closest to ideal timetabling) have even better children (closer timetabling). Eventually, an acceptably good answer is found.
One of the coolest (thought not necessarily best) computer learning projects right now is a system called NELL, which is reading the web, and learning as much as it can from the pages. It tweets everything it learns on twitter:
http://twitter.com/cmunell
Basically what it's doing is reading descriptions or facts and trying to build a mental model of how one thing relates to another -- learning that apples are a kind of fruit, for instance. When it builds up knowledge like this, it could then be used to answer questions like "how many kinds of fruit have a thick skin?", or even "I don't like things with skin. What nice meals could I make, with what's in the kitchen? Can you make it for me?"
Making computers do this is tricky, and limited, so for chess, it's easier to just feed in all the chess rules, and have the computer work out every possible move, and assign a score. Chess is relatively simple like that -- each piece only has a few possible moves. Handling more complex games, like Go, where a lot of different strategies can be used, and there are only two kinds of piece, but they can be placed anywhere on the board, with lots of different implications for each move, is more difficult for computers... but still not really "learning".
Eventually, yes, a computer will be able to do everything a human can do, and more.
Yes, it can, if pre-programmed to do that. Artificial Intelligence is being developed all over the world. Although limited, it is promising. Google for DEEP BLUECan machines and computer learn?
yes it is possible and "they" did this already.
For chess i know automatic learning is used to tune the evaluation function ( which determines how good/hopeless a position is )
No, they would have to be programmed by a human being.Can machines and computer learn?
well if it did get programed with an advanced A.I.
..only "Kitt" in the new Knight Rider...
Subscribe to:
Post Comments (Atom)
No comments:
Post a Comment