The fact of the matter is Watson can. He can devolve meaning and figure out what those doctors meant. Once you do that, you have a phenomenal opportunity to then coordinate all of that information, and take it and deep mine it.
Mine it in what ways?
Well, take a look at a particular disease like diabetes. You take all the data available globally, you anonymize it to preserve privacy, but then you crawl through it and say, “Was there some spectacular improvement that was seen in some small number of cases where people tried one strategy versus another?”
And you mine out of that data for the absolute best conceivable treatment for a patient, potentially optimized by ethnicity (if that turns out to be a factor), or if certain people have a predilection to one treatment versus another.
How far in the future do you see this ability to tap into “unstructured data” as the norm?
Well, there’s a huge cultural barrier you must overcome. There are issues of privacy that you absolutely must address, but you’re certainly not talking about outside of the 5 to 10 year window. It is coming, and conceivably within this generation or the next decade.
What about outside of Watson? Are there any other kinds of innovations IBM’s working on that will change the way we live?
You’re familiar with the phrase “real time”? That sounds great, right? Let me give you an example of when “real time” isn’t so hot, like, say, you’re stuck in traffic. Fantastic, right?
Not really. The fact is “real time” isn’t good enough. Imagine for a minute that you could collect all the data for traffic flow in a city. Not by each car alone, but exactly how many cars are in what place, at what time, and how fast they’re going.
Say for example you have a snapshot of traffic flow, and then 20 minutes later on average there’s a massive tie up that occurs at one central location. On that information, you can actually go in and start to build models for the entire city, which in fact we did in places like Singapore and London.
When you see a traffic jam, you can jump in ahead of it and basically do things like limit the access to that intersection by slowing the lights down 2 or 3 percent as cars approach it.
That doesn’t sound like much, but there’s a critical point where flow becomes chaos. It’s a very interesting mathematical theory, where the conversion from smooth flow to chaos can occur over an incredibly narrow window. It’s not that you need to eliminate all the cars, you just need to back down a relatively modest amount and the traffic jam never happens.