IBM Watson and Jeopardy - pushing language processing to the limits

I'm proud to work for IBM.

Sometimes things come along that make you really proud to work for your company and the recent news of the IBM Watson computer winning the Jeopardy gameshow challenge against two world champions (one who won 72 straight games in a row and the other who has won $3.5m on the show) is pretty cool stuff.

Jeopardy is a word game where you have to guess the word clue from a cryptic clue, and based upon how confident you feel about your guess, you bet appropriately.

There's natural language recognition, question/answer word recognition, confidence levels and a time challenge to play faster than your opponents.

So getting a computer to play this and to play at world-class level is pretty clever stuff and IBM Research have been working on it for the last 3 years or so. It's really pushing the boundaries of computer search, particularly in the field of natural language processing, and I personally find it amazing the results that have been achieved.

The main IBM Watson website has more details of the challenge and the team that put together Watson, and there's a whole host of video's of Watson and Jeopardy on Youtube - the Watson Grand Final of Jeopardy! is worth watching first, and look at the emotion on the faces of the IBM team as the computer wins.

Doubtless in a few years this will become "old hat" in the same way that computers playing Chess is now common place and was said to be "impossible", but for now the way that Watson works out from our idiosyncratic language what the question really is and then 90%+ of the time gives the right answer is frankly amazing.