In mid-February of this year, NBC broke its ratings record as millions of American viewers tuned in to a see a game of Jeopardy! The impressive turnout wasn’t due to a collective bout of nostalgia, but interest in a particular unique contestant. Watson, a super-computer engineered by a team of 24 researchers at IBM, made history as the first non-human contestant to compete on the game show and beat the defending champion, effectively earning a one million dollar prize and awakening the world to a new brand of artificial intelligence.
Steve Welch, a Distinguished Engineer and Manager of Health Informatics for IBM and a member of the team that worked on Watson, spoke at the Endeavor Summit this summer about the engineering behind Watson’s development and the implications of this innovative technology for the world.
The biggest hurdle in developing Watson, Welch said, was getting a computer to understand natural human language. Unlike computer code, equations, or the key search terms we usually use to ‘talk’ with our computers, human language is nuanced, ambiguous and contextual. Years of experience and cognitive processes go into any form of verbal human communication, making it virtually impossible for a computer to decode. The breakthrough came in the form of a new paradigm that was created by an IBM summer student. His system reconfigured the way computers process language and provided IBM with the framework that would eventually define the super-computer’s consciousness.
Watson’s ‘brain’ is composed of seven banks of processors and 2,800 cores. The team of engineers and researchers behind his development spent years inputting thousands of pages of data into Watson’s knowledge base that could be called upon when answering questions on Jeopardy (Watson was not connected to the Internet during the game.) Unlike traditional search engines, Watson weighs supporting evidence for a number of different answers before producing a response. When it reports an answer, it can also provide the evidence that led to it and a level of confidence in the findings. Just like a real human, Watson learns to trust different sources of information based on past experience.
Watson’s development marks not only an innovative leap for technology, but also for art. IBM went to great lengths to develop Watson’s humanity, bringing in a voice actor to record thousands of lines and sounds as the basis of Watson’s ‘voice,’ and hiring a generative artist to design Watson’s face. The image created is a swarming globe that has 27 different states to illustrate Watson’s various moods – when he is very confident in is answer, the globe will swarm towards the top of the screen and glow green.
A member of the audience asked Welch where Watson was now: has it retired on its earnings to a life of leisure? Of course not. For now, Watson is an IBM employee working in healthcare and finance, two industries where there is an influx of information to be processed and analyzed. In the future, IBM hopes to develop practical applications for the Watson technology that will create benefits in many different facets of society.