This week, IBM’s Watson supercomputer defeated Jeopardy‘s greatest champions, and it wasn’t even close. When all was said and done, Watson won $77,147, far more than Brad Rutter ($21,600) or 74-time champion Ken Jennings ($24,000). Its ability to dissect complex human language and return correct responses in a matter of seconds was simply too much for humanity’s best players.
A few years ago though, Watson couldn’t even answer 20% of the questions it was given correctly. And it took hours, not seconds, for Watson to process through a question.
At an intimate event in San Francisco, John Prager, one of the researchers developing Watson’s ability to answer complex questions, gave a presentation detailing the work he and his colleagues did to turn Watson into a Jeopardy champion. During his presentation and a Q&A afterwards, Prager and fellow researcher Burn Lewis revealed some key nuggets of information, such as why Watson made those odd, uneven bets during Daily Doubles (an IBM researcher thought it would be boring if Watson’s bets ended with zeros, so he added random dollar amounts for kicks) or which programming languages the researchers used to build Watson (Java and C++).
So what’s next for Watson? Prager says that the next frontier is health care; he hopes that Watson’s technology can help diagnose ailments by analyzing vast quantities of data against patient symptoms and queries.
Check out the video to get a deeper dive into the technology behind Watson. Check it out in HD if you want to read the slides.
