Watson’s Prodigy Leads Power Systems Into The Cognitive Era

November 4, 2013 Dan Burger

Your entire career has played out in the programmable era of computing. That’s not likely to be the case for young IT professionals with recently launched careers, as new trails are being blazed into cognitive computing. Leading the way is IBM‘s most famous computer, the Jeopardy!-winning supercomputer that goes by the name Watson. But Watson’s Jeopardy! fame was just the beginning of things to come. And if that’s all you know, you have some catching up to do.

“The next 20 years will change computing as we know it today,” says the person chosen to guide Watson development.

It’s been two and a half years since Watson became a household name after outperforming the best flesh-and-blood Jeopardy! contestants of all time. Shortly after that highly publicized event, Manoj Saxena was named general manager of Watson Solutions. Saxena was one of the keynote speakers at IBM’s recent Enterprise2013 conference, where he explained that most people would not recognize the new stronger, smarter, faster Watson. Based on size alone, the Jeopardy! version of Watson was a monster compared to the current iteration.

“The Jeopardy! Watson had almost 3,000 cores. It was the size of a master bedroom and is now a single Power 750 server running Watson in the network,” Saxena explained, “and it can scale to tens of thousands of users. Watson is on Power because the platform delivers I/O and performance and has flexibility to run on-premise or in the cloud. There’s tremendous capacity in a self-contained system.”

The Watson project started as R&D in 2006 and was game show ready by February 2011. Later that same year, Watson began showing what it could do for the healthcare industry, which led to agreements with prestigious organizations like Memorial Sloan-Kettering Cancer Center, the Cleveland Clinic, and MD Anderson Cancer Center, among others.

Watson can read and understand the entire medical records of a large hospital in six seconds, according to Saxena. Very impressive, especially the deep understanding of large volumes of information. The original Watson was fed more than two million pages of information, a training diet so to speak, as it bulked up the Jeopardy! challenge.

But Watson for healthcare adds to that capacity to learn through reading by including the input of doctors and researchers, some of that material in the form of unstructured data, which depends on the comprehension of human language and capability to reason. This is where the fundamental shift in the way people interact with machines becomes the “we’re not in Kansas anymore” moment.

“We believe, in time, Watson will change the way medicine is researched, taught, practiced, and paid for,” Saxena says.

When Saxena took over Watson two and a half years ago, the supercomputer was a one-trick pony designed to answer questions that fit the Jeopardy! format. Since then, Watson has grown to include 20 instances, and Saxena says that within a year there will be more than 100 Watsons, semantically tuned for specific industries such as finance, retail, healthcare, and telecom.

“Once Watson is commercialized,” Saxena says, “we will draw a line in the sand and say: Before this time, computing was nothing more than a giant calculator that could only understand zeros and ones and structured data. After Watson, there will be cognitive computers that can understand language, syntax, and semantics. Cognitive computers will reason, think, understand and learn from information.”

Watson’s ability to understand human communication and factor it into the decision making process is where programmable computing becomes cognitive computing. When Watson gives an answer, it also provides a confidence level for the answer. Decisions, more precisely defined as evidence-based hypotheses, result from weighing one answer against the other options.

Saxena refers to it as semantic discovery and contrasts that capability with the limitations of keyword searches that lack the capability to problem solve, make deductions, or sort through variables relating to the motivation of the person doing the search.

“The original Jeopardy system was a question and answer system,” he notes. “What we have now is a conversation system.”

Saxena uses a person searching for real estate information as an example. A cognitive system, integrated with a call center, would handle a question such as “Is this a good time to buy?” by processing articles and statistics on real estate availability, loan rates, what’s driving the market, and the input of cooperative brokers and other experts and deliver several answers along with evidence that substantiates the choices. It includes information that is specific to the region in which the user resides: state, county, city, zip code. Based on the information received, the caller might have related questions, which the cognitive system would be able to respond to because of its deeper understanding and semantics.

Each industry has its own semantics, so cognitive systems require fine-tuning for each industry to be able to understand what the data represents. Numerous companies are working with IBM on this for customer service and sales support, and Watson Industry Solutions for cross-industry applications began in 2012.

Cognitive computing systems like Watson are a nice fit with the pressures of how to make sense out of big data. The content analysis aspects along with evidence-based reasoning and language processing being used to identify data relationships is a key element in the goal of improved decision making.

“This journey with Watson has been one of the most meaningful things I’ve done in my life because of the impact this technology will have,” Saxena says. “Listening to the blood cancer specialists at MD Anderson tell us what they believed the technology would do was one of the most significant highlights of my career. What they were talking about is their ability to use a machine that could advise and assist them by leveraging the knowledge and power of all that is known to mankind and keeping that knowledge in the machine for the benefit of successive generations.”