David Ferrucci, the manager of the Watson project at IBM, on why he thinks Watson got the Final Jeopardy question wrong:<p>"First, the category names on Jeopardy! are tricky. The answers often do not exactly fit the category. Watson, in his training phase, learned that categories only weakly suggest the kind of answer that is expected, and, therefore, the machine downgrades their significance. The way the language was parsed provided an advantage for the humans and a disadvantage for Watson, as well. “What US city” wasn’t in the question. If it had been, Watson would have given US cities much more weight as it searched for the answer. Adding to the confusion for Watson, there are cities named Toronto in the United States and the Toronto in Canada has an American League baseball team. It probably picked up those facts from the written material it has digested. Also, the machine didn’t find much evidence to connect either city’s airport to World War II. (Chicago was a very close second on Watson’s list of possible answers.) So this is just one of those situations that’s a snap for a reasonably knowledgeable human but a true brain teaser for the machine."<p><a href="http://asmarterplanet.com/blog/2011/02/watson-on-jeopardy-day-two-the-confusion-over-an-airport-clue.html" rel="nofollow">http://asmarterplanet.com/blog/2011/02/watson-on-jeopardy-da...</a>
For all the talk of the difficulties of playing Jeopardy! due to the "nuances of natural language" and "puns and double meanings in the clues", that did not really seem to be a factor in the second round -- most of the questions were quite plainly worded with answers easily discoverable just by searching. Accordingly, Watson performed dramatically better today than yesterday, when a larger portion of the questions did have nuance and plays-on-words in the phrasing. Note too how spectacularly badly Watson performed on the Final Jeopardy! question, where nuance _did_ play a much bigger role.<p>So today, we learned that machines can push buttons faster than people, and search is a great way to find answers for trivia questions. I doubt the former is a surprise to anybody alive in the past 50 years; the latter shouldn't surprise anybody who's ever used Google.
So is there any information on how they actually implemented watson? My understanding is it's a bayesian machine learning system, but I still don't know how it parses answers, or really does its magic.<p>Also, if there is anyone who thinks silicon valley has the smartest people around, this type of stuff should change your mind. Facebook is short trousers compared to this. and it's just a tech demo.
And, as I predicted, it only came to buzzer reflex, which computers unsurprisingly excel at. On day 2 (today), Watson was only beaten to the buzzer three times when it had the correct response above its confidence threshold.
To make it a true test of brains, and remove the mechanics of button-pressing speed from the question...<p>Place all three contestants in isolation from each other.<p>All three hear the question read, and buzz-in just as they do now.<p>Allow ALL contestants who buzz in to answer the question, but do not allow them to know about their opponents' performances.<p>Record all contestants' buzz-in reaction times.<p>At the end of the game, compare only the accuracy of answers to determine the winner.<p>At the end of the game, compare buzz-in reaction times to see how thumbs fare against relays.
Just put up the latest results<p><a href="https://spreadsheets1.google.com/ccc?key=tth_jhM8vyBAuogqHllHmHQ#gid=2" rel="nofollow">https://spreadsheets1.google.com/ccc?key=tth_jhM8vyBAuogqHll...</a>
I received a few complaints when I posted the results of round #1 and it hit the homepage. You might want to change the title to something ambiguous about who won.
Part 1 of the second round on YouTube here: <a href="http://www.youtube.com/watch?v=PHhDLUVAtqU" rel="nofollow">http://www.youtube.com/watch?v=PHhDLUVAtqU</a>
I had to leave the article to search where the actual building was located because all they gave was, "suburban New York". I'm still not sure where it is.
the fact that watson has good nlp isnt nearly as impressive as the fact that it has a huge knowledge base, how the hell did it get all that knowledge, if it is just from browsing the internet by itself that makes me afraid.......very afraid