|
One thing about the Watson match is if it gets the whole text file when Trebek first starts reading the question, while the guys need to speed read it off the screen or listen along to Trebek's speech, that by itself is a 2-4 second gap which is a lot of time when deciding to hit the buzzer. I think a lot of Ken Jennings half time adjustment is realizing the couple of seconds of stall time he gets after hitting the buzzer is usually enough to finish the thought in his head, so ring in on any confidence and expect to solve it.
It is a lot different for Watson to get the question as a text file, process 2 seconds, and then halt on the automatic signal it gets when the question is done being read, than to get all of the content from speech recognition and also recognize the buzz timing.
Of course, because the other contestants can see you can't show the question on the board as well.
So to me, the fairest match is Trebek reading the questions without any visual cues, and the machine needing to parse the words from sound (which is somewhat harder) or to get each text word as Trebek starts it.
You will be surprised how often when watching the show that looking at the end of the question gives it away almost immediately, and with a search running the way Watson probably is running, getting the words in sequence would make a difference in how it can do its word association algorithm.
That said, it is still highly likely that Watson would be competitive in a match, particularly against normal contestants which have a number of questions that wouldn't make a difference whether there is a couple seconds of startup lag due to the text file advantage.
|