So if you are a Jeopardy fan, Watson (aka “Deep QA”), an IBM computing system, is taking on some Jeopardy Champions. Why? That is a good question. As you may or may not remember, almost 15 years ago IBM’s Deep Blue beat the reigning world chess champion Garry Kasparov 3.5 to 2.5. So what is the big deal about a computer taking on Jeopardy Champions—after all, most of us are familiar with modern intelligent search tools like Google, Bing, Yahoo, etc. It is not a stretch to imagine a computer finding information quickly. Well that turns out to be the easy part. So what is the big deal?
For those of you like me who don’t turn on a TV other than to watch a Netflix movie or to act as a large screen for my PS3, Jeopardy is a prime-time television game where one is given an answer on a topic and has to determine what the original question was. When the Watson research team looked into this challenge, they studied the analytics of the winners and the champions. Human Jeopardy players have to determine whether they might know the question to the answer faster than their opponents, hit a buzzer faster than their opponents, and then phrase their ‘answer’ in the form of a question. If they get it right, they get the points allocated to that answer.
Human beings work much differently than computers do. Humans assess a situation very quickly in parallel to the spoken and written word. They determine a confidence level (a direct assessment of their competitors) almost instantly. They make a quick decision on whether they know the ‘answer’ and press the buzzer. If they win the buzzer, they then have many seconds to determine their response.
Watson, being a simple Von Neumann computing machine, has to do the entire job before the buzzer. So to spell this out, Watson has to interpret the ‘question’, determine what information is available, scan the possible hundreds of potential ‘answers’ and their alternatives, create a hypothesis for the correct ‘answer’, assess a confidence level, and decide to press the buzzer. When Watson presses the buzzer, his job is done. He already knows the answer and merely converts it to audio (voice). Oh—I forgot to mention—Watson can’t be ‘connected to the internet’. That means the DeepQA database has to have all of this information available.
Now we have all waited how many seconds for Google, Bing, or Yahoo to come back with our search results and then spent countless minutes poring over the returned items trying to determine what is relevant to our query and what is not. Watson has to do all of this in microseconds. Then determine what possible ‘answers’ he can come up with (hypothesis), determine a confidence level to all the alternatives, make a choice, and ‘press’ the buzzer. I watched Watson take on some of us employees, and while he kicked our butts, he didn’t hit the buzzer first all the time. As little as six months ago Watson wasn’t at the playing capabilities of these champions. It will be a good game.
So still—why should we care? Well imagine Watson (a future version of DeepQA) having the ability to listen live to customer requests, determine what solutions they are seeking, and initiate the appropriate actions. This is the ‘holy grail’ of computing systems for all sorts of corporate and consumer needs. If Watson can do this as an ‘expert’, then it can pass the Turing test. Hopefully—no more waiting 30’ minutes to talk to someone in a call center. Interactive Voice Response (IVR) + Watson has huge potential to impact our lives.
Well, I am looking forward to seeing how Watson fares. Catch it on your local station February 14-16.