Cognitive computing delivers answers, asks new questions
The importance of IBM’s Watson computer beating two human contestants in a game of "Jeopardy!" five years ago was better demonstrated the day after the televised event when cancer researchers called to ask if Watson’s computing capabilities could be harnessed to help in that area.
That idea evolved into IBM for Oncology, one of more than 500 partnerships, including industrial and maintenance-related applications, developed around the idea of cognitive computing demonstrated by Watson on a game show can be used to help solve the complex problem of complex data in an increasingly complex world.
That’s the potential of cognitive computing, according to Rob High, vice president and chief technical officer for the IBM Watson project. As he told the annual ARC Advisory Group forum in Orlando on Feb. 10, the need for such computing power is essential to take full advantage of the knowledge humans are creating. "Cognitive computing comes down to data. There’s been an enormous growth of data," High said. "We are going to generate 2.5 exabytes of data today. That’s 2.5 billion, billion bytes. By 2020, we’re going to generate 44 zetabytes of data."
The problem with that, High noted, was not the data itself. "We can’t read it all; we only can get a small sliver of it," he said. "The disparity is between the information and our ability to consume it. We want to tap into that massive volume of information to make decisions, but to do that, we need cognitive systems." High said there were four main characteristics of cognitive systems:
- They are able to learn their behaviors through education.
- They support forms of expression that are more natural for human interaction.
- Their primary value is their expertise.
- They continue to evolve as they experience new information, new scenarios and new responses.
"Our human condition is far too complex to represent mathematically," High said.
"We don’t look up each individual word to create meaning. We derive our understanding through pattern recognition and through those signals gain meaning."
High said the questions are more daunting as humans and the cognitive systems keep learning. He posed three questions for consideration:
- How do we use cognitive systems to amplify human cognition?
- How do we make it possible to think about a problem you might not have thought about it before?
- How do you make decisions in a way you didn’t before?
If that sounds a little too mechanical, some of the new cognitive systems under development also will include contextual, linguistic and even emotional analysis, such as robots that can evaluate body language and cognitive systems that can recognize puns and innuendo. Those latter skills were important for Watson’s foray into "Jeopardy!"
"We had to interpret the context in which the question was intended, but use that context to look for information," High said. They also loaded about 200 million pages of literature into Watson to help prepare the computer to play the game. While the demonstration was a cultural phenomenon (Jeopardy! champion Ken Jennings famously wrote, "I for one welcome our new computer overlords," for his Final Jeopardy! question at the end of the second game). High said the real breakthrough was demonstrating the potential of cognitive computing. "The real breakthrough was that we could tackle problem of human reasoning," he said.
Today, that concept that evolved into 530 business partnerships, including more than 100 universities, and 80,000 developers using cognitive services on IBM’s Bluemix platform. That includes Austin, Texas-based SparkCognition, which works with aerospace companies to resolve maintenance issues more quickly based on the context of the problems faced. High said this effort has lowered maintenance costs by up to 10%.
"Cognitive systems do the research for you so you can do your thinking better," High said. "On our behalf, the cognitive system sifts through the information to find what is most important to us and helps inform our decisions. It changes the way we as humans think."
– Bob Vavra is content manager, Plant Engineering, CFE Media, firstname.lastname@example.org.
See additional stories about the ARC Industry Forum and more information about "Jeopardy!" and Watson.