Terrell Jones, the founder of Travelocity and the founding chairman of Kayak.com, let the world in on a secret yesterday: “The guy that started the online travel revolution–me–I use travel agents.”
Not to book a flight to New York or anything, but to plan his perfect vacation. “You can’t get expert travel advice on the web. You can get reviews, but you can’t get advice,” he said at an event hosted by IBM.
That could be starting to change for a wide variety of industries as IBM moves Watson–the conversant supercomputer that famously won Jeopardy in 2011–into real-world applications and seeks to usher in what CEO Ginni Rometty describes as “a new era of machine-human collaboration”–like having a computer recommend you the perfect travel destination.
Yesterday, the company said it would pour more than $1 billion and almost 2,000 staff into a New York City division that will not only help people process the exploding amounts of information around them, but also think it through and make decisions. “It may, in fact, start to re-humanize the Internet,” Mike Rhodin, the chief of the newly formed Watson Group, says.
IBM is far from the only company pursuing a market for recent advancements in artificial intelligence technology. But Watson’s unique ability to learn as it goes and IBM’s track record working with businesses and governments means that its new commercial partners–from a top cancer research institute to banks and retailers, as well as the ecosystem of startups it is now courting with a $100 million investment fund–are at the forefront of a major shift in how we use our computers.
Here are some ways that “cognitive computing” programs like Watson, which since its Jeopardy days is now faster, smarter, and smaller (it takes up the space of three stacked pizza boxes, rather than an entire bedroom), could have an impact on how people and businesses interact with information:
Siri says she’s a smart assistant, but does anyone believe her?
Someone like Travelocity’s Jones would never ask Siri where he should take his family on an adventurous vacation. But that’s the goal of the next wave of natural language assistants powered by Watson. Jones showed how Watson could quickly read through 64 million reviews, 16 million blogs, and 7,000 guides to recommend with 97% confidence that a trip to Bali was the perfect trip for him. If he gave it additional information (say he wanted more than just a beach), it spit back Punta Cana. “That’s something you can’t ask any travel site today,” he says.
Other demos show this kind of interaction in realms beyond travel, such as creating personal shopping assistants and health coaches. Fluid Retail CEO Kent Deverell showed off an “expert personal shopping” application built for North Face, where a visitor types into a simple question box: “What equipment do I need for a 14 -day camping trip?” and get in return a list of specific product categories that he can continue to query in the same way he would speak. “It is not a super search engine. It can find a needle in a haystack, but it [also] understands the haystack,” says IBM CEO Rometty.
In a world of exploding data, even experts can’t keep up. This is especially true for doctors, even specialists, who can’t possibly keep abreast of the 7,000 biomedical studies published a month, hundreds of clinical trials, and the growing pile of imaging and genomic data available for each patient.
Memorial Sloan-Kettering Cancer Center has spent the last two years with its hundreds of doctors training Watson as an oncologist’s advisor, taking the inputs of clinical data and recommending appropriate tests and treatment plans. Watson cites its evidence (or lack of evidence) and certainty level so doctors can make the judgment calls. For example, Watson turned up relevant advice about an obscure cancer tumor mutation that only 10 oncologists in the world might know off-hand. The center also wants to integrate patient preferences into the decision making process. “We feel strongly that this will change the way health care is delivered,” Craig Thompson, CEO of the cancer center, says.
One defining feature of Watson is that it learns. But can it also guide students in the learning process? The Cleveland Clinic is experimenting with that at its medical school, as students and Watson look at a case alongside each other. Students can help train and correct Watson, but they can also look at its decision trees and evidence and understand the process and data that goes into Watson’s analysis. “I call this the virtuous cycle. Who is the student and who is the teacher? Well, both,” says Thomas Graham, chief innovation officer of the Cleveland Clinic.
One new Watson product, Discovery Advisor, will allow users to not only answer questions, but find the right questions to ask. For example, a pharmaceutical company could explore chemical compounds that it has never tested before, and ask Watson questions about what’s not known. “It’s finding the white spaces … moving through the sea of data,” says Rhodin. An advantage of Watson is that it explains its reasoning, and so it allows researchers to refine their investigations in promising directions.
Despite IBM’s enthusiasm and marketing resources, there have been bumps in the road, and the Wall Street Journal reported this week that initial revenues for Watson products have been a disappointment.
One problem is that it takes a long time to teach and train the system with specialized data in each use case and sector. Mark Kris, the chief physician for Memorial Sloan-Kettering’s Watson project, says it will take years to develop all the uses they’ve foreseen for the hospital. But he expects doctors to start actually using Watson in a patient setting this year, and says he hopes other oncologists won’t have to replicate Sloan-Kettering’s training project. “This is a lot more complicated than Jeopardy,” he says.