The race to build a computer that can think like humans has been in the news recently, with IBM's computer "Watson" taking on the best human players in the American game show Jeopardy.
Since computers were invented, it has always been assumed that one day they will become smart enough to think like humans, but progress has been slow. In 1997, Deep Blue made history by defeating the world chess champion Gary Kasperov. Although mastering chess was a great feat, it only required programming the computer to understand the limited number of rules in chess. In Jeopardy, answers to questions often hide behind puns, double meanings and word play. As it was impossible to program a computer to understand all these subtleties, a different approach to computing needed to be applied.
Machine Learning is a process by which a computer learns by example. By showing computers a large number of examples, they are able to build their own rules, effectively learning now to deal with information. In Jeopardy, this involved loading all the past questions and answers from the show. When this was combined with a huge database of information from which to pull potential answers from, Watson was ready to play.
In terms of future applications, the idea that a computer can understand questions, interpret what is important in that question and return an answer from an extremely large database in seconds is very exciting. The creators of Watson are especially excited about its potential in the health care industry and have announced further research into that area (http://www-03.ibm.com/press/us/en/pressrelease/33726.wss).
To learn more about Watson and how he performed on Jeopardy, take a look at this documentary which was made before the show went to air:
http://www.youtube.com/watch?v=5Gpaf6NaUEw&feature=related
There is also an interesting Q&A with some of the engineers that worked on Watson here:
http://blog.reddit.com/2011/02/ibm-watson-research-team-answers-your.html
You can see Watson play the final day (of 3) here:
http://www.youtube.com/watch?v=xm8iUjzgPTg
Since computers were invented, it has always been assumed that one day they will become smart enough to think like humans, but progress has been slow. In 1997, Deep Blue made history by defeating the world chess champion Gary Kasperov. Although mastering chess was a great feat, it only required programming the computer to understand the limited number of rules in chess. In Jeopardy, answers to questions often hide behind puns, double meanings and word play. As it was impossible to program a computer to understand all these subtleties, a different approach to computing needed to be applied.
Machine Learning is a process by which a computer learns by example. By showing computers a large number of examples, they are able to build their own rules, effectively learning now to deal with information. In Jeopardy, this involved loading all the past questions and answers from the show. When this was combined with a huge database of information from which to pull potential answers from, Watson was ready to play.
In terms of future applications, the idea that a computer can understand questions, interpret what is important in that question and return an answer from an extremely large database in seconds is very exciting. The creators of Watson are especially excited about its potential in the health care industry and have announced further research into that area (http://www-03.ibm.com/press/us/en/pressrelease/33726.wss).
To learn more about Watson and how he performed on Jeopardy, take a look at this documentary which was made before the show went to air:
http://www.youtube.com/watch?v=5Gpaf6NaUEw&feature=related
There is also an interesting Q&A with some of the engineers that worked on Watson here:
http://blog.reddit.com/2011/02/ibm-watson-research-team-answers-your.html
You can see Watson play the final day (of 3) here:
http://www.youtube.com/watch?v=xm8iUjzgPTg
No comments:
Post a Comment