by Peter High, published on Forbes
1-18-2016
Perhaps nothing signaled the arrival of artificial intelligence quite like January 14, 2011, when IBM Watson defeated two legendary Jeopardy! champions, Ken Jennings and Brad Rutter. At last, the computers were smarter than us. Since then, Watson has developed into a growth engine for IBM.
For just over two years, Michael Rhodin has led IBM Watson, having come to that role after a four year stint as the head of the Software Solutions Group at the company. IBM has already found applications for Watson across seventeen industries in over thirty five countries, but as Rhodin notes in this interview, the company is just getting started.
(To listen to an unabridged audio version of this interview, please click this link. This is the third article in a series on leaders in artificial intelligence. To read future articles in the series, please click the “Follow” link above.)
Peter High: Many people will remember that Watson was developed initially to compete on Jeopardy! versus two of the biggest Jeopardy!champions of all time, and it won a million-dollar prize in the process. Then a bit more than a year ago, IBM made the decision to create a business called the IBM Watson Group, which you now lead. The company invested a billion dollars (quite a bit more than the winnings from Jeopardy!) to get the business off the ground. Was there always a desire to turn Watson into a business line?
Mike Rhodin: If you look at what we were trying to do, this was an IBM Research Grand Challenge going after– from a technology viewpoint– something in the AI world that was known as the “open domain deep Q&A problem”. It was a problem that was originally postulated based on a comment by [John] von Neumann in the late 1940′s. At that point in time, after his architecture became the basis for modern computing (the binary system that all computers today are built on), he made a comment in the late ‘40s that said someday computers will be able to answer any question. For the next sixty years, computer scientists were trying to figure out how to do what he said, which is how to build a system that could answer questions on a broad base of domain knowledge. That was the real impetus for the Watson system coming out of IBM Research.
Along the way, we decided that it was going to be necessary to have a proof point that you could answer questions in an open domain. The Jeopardy!quiz show ended up being a great example of something that had a broad topic base, lots of different types of questions, tricks, use of natural language in interesting ways, and it became a great demonstration of the technology and the breakthroughs that the IBM Research team made. So that was what led up to Jeopardy! in the first place.
After the Jeopardy! match, we were thrilled with the outcome. It was not a foregone conclusion. With any of the probabilistic systems, there is a level of chance in everything, but we knew we had a system that we thought would show well in the game show. We were pleased to be able to come out with a win against two incredibly smart leading champions of the game.
We started a period of what I would consider in-market experimentation for the two years after the Jeopardy! match. We started working with a handful of companies that had approached us that wanted to start to experiment with the technology–not to play Jeopardy!, but to use the underlying technology to start to solve problems. The preponderance of them were in the healthcare field. Industry luminaries like Memorial Sloan-Kettering or MD Anderson in cancer, or in general medicine, the Cleveland Clinic. They all started partnering with us to explore how we could take what was inside the Jeopardy! system and morph it into things that could be used in the healthcare profession. We did that for a couple years, and that gave us the confidence that led to the announcement in 2014 of the commercialization project and Watson Group.
When you think about what a doctor does every day, they gather evidence about a patient, they use that evidence to build a level of confidence in a set of hypotheses that could lead to a diagnosis, and then they make a treatment based on that. They were looking for systems that could start to help them with a particular problem that has occurred in the healthcare industry, that is, the amount of information being produced: the amount of new research and publications has outstripped the ability of doctors to keep up. In 2015 alone, we will produce something around seven hundred thousand new reference documents in medicine. I am pretty sure most doctors do not have time to read all of them. They are focused on how we could start to use what we demonstrated in the Jeopardy! match in the pursuit of helping them understand all of the information that is being published in their profession, and how to do it in a way that could help them create better outcomes for their patients. That was one of the first clues that the AI world was ready to be woken up– we have been a little bit dormant for decades — and that the systems were ready to start to move into prime time.
That was how we got to the launch last year. It became a pretty big point for us because we had decided that not only could this technology be used in health care, which we still believe is a huge opportunity, but we also recognize that it had applicability across pretty much any industry we could see. The way we thought of it was that any profession within any industry where the amount of information being produced has surpassed the ability of the humans in those professions to consume it. That is a lot of professions these days with the amount of information being produced. That part became pretty clear to us.
The second thing that was a key decision about the launch of the commercial project was the creation of an open ecosystem: we would open up the APIs on platforms so that startups could get access to the technology and start to build out businesses on top of it. When we launched the Watson Group we had our early adopter customers that we had been working with, but we also had a small number of startups that we had exposed to the technology in advance of the launch so that they could be ready to stand up with us and talk about we were doing. Since then, the ecosystem project has taken off. We have hundreds of companies building on the platform, over one hundred now in-market with commercial solutions. Another four hundred or so are under development behind that, and they will be coming out over the next year or so. That part is taking off as well.
