I majored in English, a notoriously difficult skill set to commercialize. (But since nothing will come of nothing, I do try to shoehorn in gratuitous Shakespeare references on this blog wherever possible.) And so I can sympathize with the plight of IBM’s Watson. You train and train and train for one thing: to be “Jeopardy!” champion of the world. And then what?
IBM’s Bernie Meyerson, the company’s vice president of innovation, tells Bloomberg that Watson’s future home may not be on your television, in front of Alex Trebek, but rather in your pocket. In fact, Meyerson thinks Watson could be the germ of a Siri-like assistant that out-Siri’s Siri (see “Getting Your Phone to Give You a Hand”).
Watson’s already making decent money, actually, in the enterprise market (that is, working for businesses, rather than consumers). Like a B.A. with a humanities degree flung into the real world, Watson has set aside its useless command of history and culture and has settled for a job in consulting, doing remunerative calculations of various sorts for Citigroup Inc. and for WellPoint Inc. (financial data for the former, cancer data for the latter). The main reason Watson isn’t already in our pockets, said Meyerson, is that he’s simply too smart–it takes more power to tap his brain than our smartphone batteries can currently muster. But the amount of power required is “dropping down like a stone,” said Meyerson. “One day, you will have ready access to an incredible engine with a world knowledge base.”
It appears that IBM doesn’t have its sights on the consumer market for now, instead wanting to roll out Watson’s services to specific corporate customers. (Watson’s “brain” is 10 racks of IBM servers in Yorktown Heights, New York; the idea isn’t to replicate that brain on each phone, but rather to make an app that would draw upon that brain’s computational power.) The idea is for Watson not to be a generalized personal assistant like Siri, but to be a specialized one–the way real personal assistants often gain expertise in their bosses’ line of work. Watson supposedly will have deep enough an understanding of oncology in the next few years to even give advice to doctors on prescriptions and diagnoses, per Bloomberg. Watson would essentially grow as a handy tool for businesses that already contract (or will newly contract) IBM’s business analytics services (see “With Watson, IBM Seeks to Sell Medical Knowledge”).
IBM told Bloomberg that some of the features we associate with Siri, like voice recognition and language processing, will come easily to Watson 2.0. But I’m not so sure that all the skills of Siri are so easily transferred. Siri’s main trick is her ability to understand just what a user is asking. Watson’s main ability has been to consume and retrieve vast amounts of information–an “ability to evaluate the relative merits of the thousands of candidate answers,” as a TR report put it back when Watson became the world’s top TV trivia pro. Siri’s great trick is inferring meaning; Watson’s great trick is retrieving information. The two skills are related, but not the same. What would probably benefit consumers and businesses most would be some offspring of Siri and Watson. But with the tech world being what it is, good luck getting those two to mate.
When designing an embedded system choosing which tools to use often comes down to building a custom solution or buying off-the-shelf tools.