Nearly three years after IBM's cognitive computing system Watson beat two human contestants to win Jeopardy!, the company is still working out how to make a reliable return on its investment in the quiz show champion.
IBM's Watson business unit, set up in the wake of its 2011 Jeopardy win to commercialise the technology, has so far brought in less than $100m in revenue, the Wall Street Journal reported on Monday — or around 100 times the prize money Watson managed to win on Jeopardy!.
According to the paper, which cited a transcript of an IBM conference call with the company's CEO Ginni Rometty, IBM hasn't "figured out how to generate a reliable revenue stream" from the product.
Must read: IBM Watson: How the Jeopardy-winning supercomputer was born, and what it wants to do next
Since the creation of the Watson business unit, IBM has chiefly been targeting the healthcare and financial services industries. In 2012, IBM signed up US healthcare providers Wellpoint and the Memorial Sloan-Kettering Cancer Centre as pilot customers, putting the system to work in cancer care, while Citigroup also said it would examine whether Watson could be used at the bank for customer service.
IBM has since begun offering Watson as a general analytics product for businesses' customer service, marketing and sales functions.
Manoj Saxena, the head of IBM's Watson unit, said in an interview with ZDNet last year that while healthcare had been the starting point for commercialising Watson when the unit was set up, any "information-intensive" industry was considered fair game.
"The healthcare discussions were already underway [before the creation of the Watson business unit] — we didn't know where in healthcare, but we knew healthcare was a good domain. One of the first things I had to do was figure out which markets do we take Watson to.
"What became apparent to me after we did some work was that information intensive industries like healthcare, insurance, banking, telecoms where there is a tremendous amount of semi-structured and unstructured data — not just transactions like retail transactions or supply chain transactions but information like doctors' notes, prospectuses on a stock purchase agreement or insurance policies — where there is a lot of human language that knowledge workers have to read and understand, are a great target for Watson, because until Watson, machines could not understand human language and unstructured information."
Watson is designed to query huge amounts of structured and unstructured data to produce useful answers — for example, in the case of cancer care, studying a patient's symptoms, diagnosis and preferences to suggest treatment.
The system is based on IBM's massively parallel software architecture DeepQA, DeepQA works out what the question is asking, then works out some possible answers based on the information it has to hand, creating a thread for each. Every thread uses hundreds of algorithms to study the evidence, looking at factors including what the information says, what type of information it is, its reliability, and how likely it is to be relevant, then creating an individual weighting based on what Watson has previously learned about how likely they are to be right. It then generated a ranked list of answers, with evidence for each of its options.
While the revenue from Watson to date remains modest, IBM still has high hopes for the technology: Rometty expects Watson to generate $10bn in annual revenue within 10 years, according to the WSJ.
IBM declined to comment on this story.