Monday, February 14, 2011

Jeopardy! vs Watson - Part 1: Watson holds his own

Well the results are in after Watson took on the greatest human Jeopardy! champions of all time in a first-ever computer versus human edition of the game.

At the end of night 1 of play, here are the results:
Ken Jennings: $2,000
Watson: $5,000
Brad Rudder: $5,000

That's right. One human is tied with the computer, with the other human trailing.

Watson was very impressive, but certainly not perfect after the first game. In one example of a glitch, Ken Jennings buzzed in with an incorrect answer and Watson buzzed in to get it correct. The computer ended up giving the exact same wrong answer as Jennings. Watson can not benefit from others' wrong answers, which impedes any possible process of elimination.

The computer also had a tough time distinguishing negation. One question asked for the "first non-dairy coffee additive." Watson answered milk. The answer made complete sense without the word "non-" in front of dairy.

The computer also seemed to have a tough time making progress in some pop culture categories. One question that boasted a correct answer of "Voldemort" from Harry Potter stumped Watson entirely.

The show also included details in how Watson was created and interviews with the experts who created him. They discussed the greatest challenge in being similar words and word association. One example, Watson has a hard time distinguishing the difference between "John McCain running for president" and "an athlete running a race." I guess context would be the issue at point here.

Watson continues on Jeopardy! tomorrow night in part 2 of the tournament.

3 comments:

  1. this new post is showing up in the feed to the class blog--I don't know why

    ReplyDelete
  2. never mind--working; a better experiment--let a human, Watson, and a crowd compete

    ReplyDelete
  3. do you think this could lead to new types of game shows where humans and computer will compete regularly? or what about humans building computers to compete against other computers? this could open many opportunities for new television.

    ReplyDelete