After Watson won the Jeopardy! game show in 2011, IBM received calls from doctors. They were swamped by the amount of medical literature and research being published but wanted to make sure they gave patients the most effective treatment based on current wisdom.

Could Watson help?

Rob High, chief technology officer for IBM Watson, calculated that if a doctor read every article published every week, just in their field, it would take 160 hours a week. “There are only 168 hours in a week,” he says.

Working with the Memorial Sloan Kettering Cancer Centre, IBM developed Watson Oncology, which is starting to be deployed internationally. It provides support to doctors as they diagnose and treat patients, analysing vast swaths of research and published papers and leveraging micro-segmentation to tailor treatment recommendations for individual patients.

“Cognitive computing is about amplifying human cognition. They don’t do your thinking for you, they do your research for you so you can do your job better; that could be as mundane as asking a question or as profound as recommending the right treatment,” says High.

Understand the nuance

In a corporate setting asking a question and getting an accurate and contextually relevant answer can also prove powerful.

“Imagine you’re in a conference room, you’re trying to make a decision about what new product to put out or what market to invest in. If you had a cognitive system at your disposal, you could just say, ‘What were the sales of smartphones in Queensland in the fourth quarter of last year?’”

The cognitive platform would hear the question, understand the nuance, and reply without anyone having to fire up a laptop or search a database, allowing a much more informed decision to be taken on the spot.

While IBM is developing solutions for particular market segments using Watson, it’s also made the platform available to independent developers who can use the services and engines to build their own systems.

In the future, all cognitive systems are likely to be much richer and more valuable. High says quantum computing could affect the speed of processing while the advent of neuromorphic chips, which are available in experimental quantities, is likely to deliver significant gains.

“Cognitive computing is based on deep learning neural nets which have some degree of similarity with how the human brain functions. But the neural nets are very basic compared with human neural systems: we currently use one sort of neuron, one type of synapse; the human brain has at least 150 different types of neuron,” he says.

Degree of probability

In time, systems currently tuned for inductive reasoning – where the answer to a question exists somewhere in the knowledge base and can be served  up with a degree of probability – could be extended to tackle more deductive reasoning. This will have particular value in healthcare and finance and, further out, High says, systems could be tuned for abductive reasoning. “These are all forms of reasoning intended to help humans do their jobs.”

And in time will the computers actually take those jobs? “We have to think about the future, but I don’t think we are at any risk of this technology taking on a life of its own.”

This story was produced in partnership with IBM.