Jones Beene <jone...@pacbell.net> wrote: > An ability to learn from an interactive network is the key - even if one > never gets out of cyberspace. Because the bird-brain-PC is essentially > tireless, working 24/7 it will be able to surpass the ability of the human > model for many tasks when given the chance... > There are relevant examples of this already.
The Watson computer "learns" from the Internet. I read somewhere it won at Jeopardy in part because it was fed the entire corpus of Wikipedia. That seems like cruel and unusual punishment for a sentient intelligent entity, but I guess we don't need to feel sorry for Watson. The AlphaGo program that beat the human grandmaster go player did not mine the internet as a way to learn. But it did "mine" two other sources. It was given 30 million moves from human game (Sci. Am.). It also played millions of go matches with itself, first taking one side, then the other. (I think it was millions.) It was learning from experience. A computer recently beat some of the world's best poker players. In other words, an AI program learned to lie, or bluff. Yet another milestone, and something that people have long predicted a computer could never do. http://www.independent.co.uk/life-style/gadgets-and-tech/news/ai-poker-win-tournament-software-beats-pro-players-victory-a7555791.html - Jed