The evening concluded with Bhavani Iyer taking a few questions from the audience on the Question and Answer Technology Watson uses. Bhavani Iyer is responsible for achieving the speed in the DeepQA technology that makes the system competitive in a live Jeopardy match. One audience member asked,"How is Watson processing the Jeopardy questions? " Bhavani explained that Watson considers internal representation of analyzed content of about the equivalent of 200 million pages when attempting to answer a clue. Bhavani explained that Watson's DeepQA processing is both CPU and memory intensive. Initially, Watson's computation took 2 hours to answer a question on a single CPU. By scaling out Watson's computation over 2,880 POWER7 processing cores and optimizing the computation and access to its data, Watson is able to answer most questions in about 3 seconds enabling the system to compete on Jeopardy. This how Watson can search his machine brain, just like a human processes a question through the brain pathways looking for an answer. Another question for Bhavani " What was the underlying framework within Watson used to scale things?" "We used Apache UIMA," said Bhavani. "This is a technology developed in IBM and used as the framework for the processing. This is Opensource and can be downloaded from www.uima.apache.org"
IBM Smarter Computing Blog
with Tags: lawrence X
Mary Hall 0600011C92 email@example.com Tags:  research rick jeopardy 7 power lawrence iyer ibm watson bhavani 3,819 Visits
Pictured: IBM Watson in Action: Buzzing in with the right answer to take the lead
Some of you may have heard that on Wednesday night, the Supercomputer known as IBM Watson defeated the two biggest "Jeopardy" winners of all time: Ken Jennings and Brad Rutter. IBM Watson finished the Jeopardy competition on Wednesday night with $77,147 in winnings over Jennings' $24,000 and Rutter's $21,600. Whether you turn on the TV, open the newspaper or log-on to Twitter, you'll find news of Watson's victory. Watson is sweeping the media with headlines shouting that in the battle of man versus machine, the machine won. Well, that's not quite true. Yes, Watson is a machine, but the Supercomputer is a machine created by humans. The brains behind Watson are the researchers and IBM technologists who developed this creation. Wednesday night at the special Jeopardy screening viewing at the Landmark Theatre in Los Angeles, I had the pleasure of meeting some of the brains behind Watson. Two brilliant IBM Researchers joined a full house of IBM customers at the event. Just as fascinating as watching IBM Watson play Jeopardy, was the live question and answer session that followed the Jeopardy TV show. Rick Lawrence, Manager, Machine Learning at IBM Research and Bhavani Iyer, Software Developer, IBM Research held the interest of the capacity crowd for almost an hour after the Jeopardy show ended. The human brains behind Watson shared the fascinating back story of how Watson was built and programmed for Jeopardy success.
Pictured: IBM's Bhavani Iyer and Rick Lawrence at the Los Angeles Jeopardy viewing party for IBM Watson.
Both Rick and Bhavani were amazed by the audience's enthusiastic response to Watson's Jeopardy win. "Watson sparks the audience's imagination and everyone is fascinated by Watson whether they are technical or not," Bhavani told me after the event. "This is because Waston changes the way humans interact with computers," says Bhavani. "No more searching by one keyword for an answer, now you can just ask a question." I have a feeling, Watson will be answering a lot more questions. Think of the possibilities for enhancing decision making in scenarios like health care. The possibilities are endless.