I spent yesterday afternoon with colleagues at IBM Watson in Manhattan’s East Village – only 30 miles from the company’s Armonk headquarters, but a world apart. (No ties, scribbling on the walls, etc.)

What began as a parlor game – can a computer play Jeopardy? – has become a desperately serious enterprise, on which IBM IBM management is betting the future of the company. “Our moon shot,” says CEO Ginni Rometty.

Watson is an attempt to leapfrog the current generation of big data services, which rely largely on structured data and rules-based computer calculations, to create a service that ingests vast amounts of information and images from disparate sources, learns from that data, responds to natural language questions, and provides real insights. The group is making its first big bet in health care, with hopes of vastly improving the ability of clinicians to diagnose illnesses, and greatly accelerating the ability of companies to discover and test new treatments. It will move from there to finance, education and the broader “internet of things.” It is also creating APIs for small businesses. While there, we met Amy Gross, who is using Watson to build a service that will help shoppers find wines that suit their tastes.

“This is the most incredible thing I’ve ever seen in my career, “ says John Kelly, who leads the Watson effort. He currently employs more than 700 PhD scientists, and that number is growing fast. Because scale is so important in the effort, Kelly believes there are probably only three companies that can compete – IBM, Google GOOG and Microsoft MSFT . And he clearly likes his company’s odds.

He also enjoys strong support from the boss. IBM has racked up thirteen straight quarters of sales declines as its traditional businesses falter, but Rometty says it is managing for the long haul. Says Kelly: “I’ve gotten everything I’ve asked for.”

Subscribe to CEO Daily, Fortune’s daily newsletter on the top business news of the day.