Dear Interweb, perhaps you can help me answer a question:
As you probably heard already, IBM built software (called Watson) to play Jeopardy – a well known game show
testing round recently where Watson appears to have won easily, which triggered a lot of articles about this. Most of them discuss the fact that the natural joke/pun filled-language is hard to understand, but I couldn't find anyone answering the most important question: How does Watson get the questions, and with what timing?
I don't think they also tackled audio recognition, so I assume Watson gets the question in text form. But when? The whole question at once, as soon as it is shown? That would put the computer at a enormous advantage, as it can
read much faster than any human. – If they'll put a static delay in, there might be cases where Watson hasn't seen the question before a human buzzes in. The only remotely fair thing that I can imagine is delaying each word separately for a short time which would mimic the show host Alex Trebek reading the question out loud.
I think this is pretty important, as the advantage of getting the whole question immediately is pretty big. What do you think?