Things are a little different.
I don’t know if Watson is getting slower or Ken Jennings and Brad Rutter have gotten better at buzzing in and anticipating the timing. Whichever it is, the humans are faring better than in the beginning.
It might be that Watson is slower on some types of questions than others and that the level of certainty may not be there for it to buzz in immediately. If it’s human adaptation, Brad got it figured out first but Ken came on strong.
I can see definite problems with how Watson analyzes the answers and falls short in certain ways. It has problems with some phrasing and structure – certain compound questions, and there are problems with associating definitions with words and selecting the right name or proper noun to settle on a response.
I don’t know if the test questions weren’t varied enough or if there is a problem that couldn’t be solved in time, but there are bugs in the analysis algorithms that I see as human in nature rather than hardware issues. Since some of the answers were properly parsed and others weren’t, there’s a depth to the analysis that is being inconsistently applied.
I wasn’t surprised that Watson won, but after watching I think that I could set up a Jeopardy! game that Watson lost. It depends on the distribution of question types.
Still, it was interesting. It was really just a 90 minute commercial for the IBM System 750 computers, but it was interesting. I’m sure there will be improvements.
Though I do wonder why Watson picked Toronto in the US Cities category
No comments:
Post a Comment