Dr. Mark Kris is among the top lung cancer specialists in the world. As chief of thoracic oncology at Memorial Sloan-Kettering (MSK) Cancer Center in New York City, he has been diagnosing and treating patients for more than 30 years. But even he is overwhelmed by the massive amount of information that goes into figuring out which drugs to give his patients -- and the relatively crude tools he has to decipher that data. "This is the standard for treatment today," he says, passing me a well-worn printout of the 2013 treatment guidelines in his office. We choose a cancer type. A paragraph of instructions says to pair two drugs from a list of 16. "Do the math," he says. It means more than 100 possible combinations. "How do you figure out which ones are the best?"

It's a huge problem. More than 230,000 Americans will be diagnosed with lung cancer this year. Almost all of them will receive chemotherapy. As crude as the existing guidelines are, says Kris, they won't be followed more than half the time. If we bumped up adherence by just 10% to 20%, he says, as many as 30,000 people might live longer. Never mind curing cancer -- shouldn't we be able to get the best available combinations of medications to sick people now?

That's the question that led Kris to IBM. He saw that more information was not the answer. What doctors needed was a better brain -- one that could instantly vacuum up facts, draw deeper connections between data points, and remember everything. They needed Watson.