Projects like these will only help Watson expand its reach into new corners of the healthcare industry and beyond, said Gold.

"When you look at the healthcare continuum, it's taking us outside of the institution and the clinical practice," he said – powering everything from large-scale genomic research projects to personalized wellness and prevention apps.

Look at Welltok, which raised $22.1 million in Series C funding from IBM and others – the first direct investment from the Watson Group – this past February. (See TK, page TK.) Its CaféWell platform seeks to bring organization and derive insight from the myriad wellness apps, health management programs and tracking devices to drive better outcomes.

"We started with institutions, moved out to application providers and, no question, throughout the course of this year we'll go out to new forms of health providers, payers, distributors," said Gold. "We can start to see how the ecosystem is starting to come together."

The technology today is faster: "24 times what it was three years ago," he said. It's smaller, too, by a factor 90 percent compared to the original technology's footprint.

"What that's done is it's opened up the possibilities of where we can start to apply it," said Gold. "We can take this into all new areas of healthcare where, before, the infrastructure requirements would have been prohibitive."

Most users, be they doctors, patients or clinicians, are well-used to programmatic computing, said Gold. "We understand it well: logic, rules, structured data. It's been around since the '50s. We get it."

With Watson, – learning as the years go on, evolving and applying its own machine algorithms to get smarter and more intuitive – "we almost have to change the way we think about computing," he said.

Especially now, as we dive even deeper into the roiling universe of big data.

"Generally, I think healthcare does a pretty good job of using structured data – systems of measurement, and metering and monitoring," said Gold. "Where healthcare doesn't do a good job is unstructured data: clinical notes, dictated into the EMR and never transcribed. Patient input. Studies, periodicals, things that have been annotated, regulatory summations."

But there's even more than that: "We think of unstructured today as text. But there's natural curiosity today about X-rays and CT scans and MRIs. Other forms of unstructured. One of the capabilities that Watson will have – it's in the lab – is the ability to read an X-ray. What would that mean for better diagnostics, for efficiency, for cost implications?"

A recent study in the Journal of the American Medical Association found that whole-genome sequencing can cost more than $17,000 per person, and will take as many as 100 hours to derive meaningful analysis from the data.

"The gist of it is we found that the results are generally not clinically acceptable," said Frederick Dewey, lead author of the analysis. "It's a relatively sobering thought, and there are tough hurdles to get over before this is common."

On March 19, the New York Genome Center announced that it would be collaborating with IBM to work on just such a challenge, testing a Watson prototype designed genomic research on personalized oncology.

Aimed specifically at glioblastoma, an aggressive form of brain cancer, Watson will target the complex research process, looking for patterns in genome sequencing and other healthcare data, such as medical journals and drug databases.

"Since the human genome was first mapped more than a decade ago, we've made tremendous progress in understanding the genetic drivers of disease. The real challenge before us is how to make sense of massive quantities of genetic data and translate them into better treatments for patients," said Robert Darnell, MD, president and scientific director of the New York Genome Center, in a press statement.

"Applying the cognitive computing power of Watson is going to revolutionize genomics and accelerate the opportunity to improve outcomes for patients with deadly diseases by providing personalized treatment," he said.

"Originally, Watson was sitting around with three or four capabilities: natural language, hypothesis generation, evaluation and learning," said Gold.

That was only three years ago. "Today," he said. "We're starting to get into reasoning, visualization and exploration."

What's next?

"We've half joked about the idea of sensory evolution – can it hear, can it speak, can it hear, can it smell? The answer is that, yes, forms of that will proliferate, absolutely," said Gold.

"Watson is a learning system," he said. "I always tell people, 'Don't get all hung up on where it is today. It's going to be different tomorrow.'"