This account is pending registration confirmation. Please click on the link within the confirmation email previously sent you to complete registration. Need a new registration confirmation email? Click here
NEW YORK (
IBM(IBM - Get Report) and its Watson computer system pushed us one step closer to technological singularity Wednesday evening.
In the second and final
Jeopardy match that pitted two of the game show's most winningest champions against IBM's natural language-grasping computer, the computer won -- easily.
"Who is Bram Stoker? (I, for one, welcome our new computer overlords,)" wrote Ken Jennings for his Final
Jeopardy answer. Jennings, who holds the record for the longest winning streak on the show, came in second with a two-game total of $24,000, Brad Rutter at third with $21,600 and Watson, who doesn't have the ability to display any emotion on his brightly-lit avatar, raked in $77,147.
"What have we learned about Watson?" asked Alex Trebek, host of the
Sony(SNE - Get Report) Pictures-produced show, while opening Wednesday's episode. "Watson is fast, Watson is capable of some weird wages and Toronto is a U.S. city."
Trebek was referring to Watson's second-night gaffe of a Final
Jeopardy answer, in which Watson answered "Toronto???" for a clue titled "U.S. Cities."
Watson had no such trip-ups Wednesday. He opened the round in much the same way that he dominated Tuesday's segment, continually beating Jennings and Rutter to the buzzer, which had reporters wondering if Watson holds some sort of advantage in ringing in.
IBM researchers, who hosted a screening of the show for reporters in New York City and San Jose, Calif., Wednesday, say no. Humans, they say, have the advantage of being able to interpret the clue before Trebek finishes speaking -- Watson doesn't. And while humans must wait to see a light that signifies it's OK to buzz in, Watson waits for an "enable signal" that's sent to him via wire.
If Watson is highly confident with an answer, he can buzz in as quickly as 10 milliseconds. "When it comes to buzzing in, humans have the advantage -- they just aren't consistent," said David Ferrucci, the project manager for Watson.
To explain the times when Rutter or Jennings beat Watson to the button, Ferrucci noted that a look at Watson's answer confidence levels -- which
Jeopardy displayed for each question -- showed that he wasn't happy enough with his top-choice answer to ring in. "It wants to get beaten when it doesn't have high confidence -- it doesn't want to look stupid."