How is IBM deploying Watson these days?

IBM on Tuesday revealed details of how several customers are putting Watson to work, showing that cognitive computing has garnered at least an initial interest among different sorts of businesses. Naming customers also helps other businesses feel more at ease about trying the new technology.

In Australia, the ANZ bank will allow its financial planners to use the Watson Engagement Advisor to help answer customer questions. The idea is that the bank can then better understand what questions are being asked, so they can be answered more quickly.

Also in Australia, Deakin University will use Watson to answer questions from the school’s 50,000 students, by way of Web and mobile interfaces. The questions might include queries about campus activities or where a particular building is located. The service will be drawn from a vast repository of school materials, such as presentations, brochures and online materials.

In Thailand, the Bumrungrad International Hospital will use a Watson service to let its doctors plan the most effective treatments for each cancer patient, based on the patient’s profile as well as on published research. The hospital will leverage research work IBM did with the Memorial Sloan Kettering Cancer Center to customize Watson for oncology research.

In Cape Town, South Africa, Metropolitan Health medical insurance company will be using Watson to help provide medical advice for the company’s 3 million customers.

Watson is also being used by IBM partners and startups as the basis for new services.

Using Watson, Travelocity co-founder Terry Jones has launched a new service called WayBlazer, which can offer travel advice via a natural language interface. The Austin Convention and Visitors Bureau is testing the WayBlazer app to see if it can increase convention and hotel bookings.

Veterinarian service provider LifeLearn of Guelph, Canada, is using Watson as the basis of a new mobile app called LifeLearn Sofie, which provides a way for animal doctors to research different treatment options. The Animal Medical Center in New York is currently testing that app.

Watson is also being incorporated into other third-party apps serving retailers, IT security and help desk managers, nonprofit fund-raisers, and the health care industry.

There is more information here.


Impressive sure, but it's no different than "heuristic rules" in software expert systems, like used for "Natural Language" searches. No biggie. What would really impress me is to teach Watson how to play a game of chess, after explaining the rules.

Watson is vastly overrated in my opinion (the organization I work for spent hundreds of thousands on it). IBM really knew how to pull off a marketing stunt by putting it on Jeopardy.

IBM pioneered the use of Big Demo.

This article on machine translation in 1954 was good at highlighting why actual reduction to practice, as it is known in the patent law field, is not always good at showing an invention really works. We need milestones too (words per second, random text, etc). That's why I'm in favor of extending the term of protection for patents and offering a prize fund if a group hits a certain milestone. For example, take fusion: if you can get more energy out than in, for a sustainable time, say 10 seconds, with very little radiation byproducts, then you would get a patent for 50 years (not just 20 like today,which is too short for pioneering inventions and too long for simple design change / incremental improvement / computer software inventions, which should be limited to a few years at best) and a cash prize from the government. This would push the envelope on Solow type 'exogenous' growth from innovation, and make it 'endogenous'. It's a crying shame that governments have not done this already. Some prizes are offered in the UK, in the Millennium project as AlexT has said, but it's not systematic.

If there are significant engineering obstacles to making a breakthrough technology practical, the solutions to those obstacles are usually themselves patentable.

@Cyrus- I agree, and we could talk about blocking patents too, but I'm talking about a hypothetical improved patent system for the future. I think there should be a non-transferable license to practice an invention that you made work, even if somebody else has a paper patent, Jerome Lemelson style. In other news: the Japanese Blue Light LED laser inventor, who made groundbreaking IP law in Japan, won the Nobel Prize today.

Don't forget "Chef Watson"

I heard the Peruvian Poutine was a big hit at SXSW

I'm skeptical if these are "customers" of IBM Watson. Are they being actually charged for use? Or is this typical demand stimulation. Customer pull or producer push?

I could see IBM deciding to charge a low rate to a university if they think it is strategic, but customers such as banks and insurance companies are definitely paying up big time.

That said, I work for a university, and we paid (a lot) for Watson.

Ok, often I've seen situations in which a corporate anyways buys some million dollars worth of IBM equipment every year and then IBM throws in a new product to be able to say that some big name corporate is using it. Sometimes it's shown as a sale but the actual value is offset by equal discounts elsewhere. It's a win win for both sides from the PR perspective.

Any case, I'm curious: What's your university using it for?

Yes, would love to know what an educational institution is using it for. We were told it would be great to throw all our SIS and Course data at it to see what it crunches up and spits back out.

Research using free text medical records.

It's surprising that your university paid "a lot". At this stage of the adoption curve, IBM is hungry for any successful application and will likely settle for a nominal payment as long as the customer is credible, willing to issue press releases, and promises not to disclose the terms. This is one of those situations where you can grind IBM hard.

Are some of them just normal expert systems cashing in on the name of Watson?

Watson is, more or less, an expert system tied to some natural language processing capabilities. Useful results in a specific domain involve a huge amount of training and tweaking.

For all the publicity it's been getting, has IBM exposed Watson on the net anywhere? Is it possible for a lay user to play around the stock expert system. Say, something like the Jeopardy app?

This has been making the rounds today:

Thirty odd years ago I knew some academics who worked on what they called Artificial Intelligence. I ignored their propaganda and formed the impression that AI's usefulness was probably several decades away. It was a bit like fusion power stations, except that it was obvious that they'd always be forty years away.

God, I am bored out of my freaking mind!!!!

Poor dear, you were expecting a singularity and got a position as a station in a service queue.

So a university needs an expensive computer to keep track of its offerings? I can understand treatment options, which change and are naturally complex. How about cutting the complexity instead of cementing it in by some new toy?

"Veterinarian service provider LifeLearn of Guelph, Canada, is using Watson as the basis of a new mobile app called LifeLearn Sofie, which provides a way for animal doctors to research different treatment options. The Animal Medical Center in New York is currently testing that app."

A sentence like this should send the writer to literary Hades. Guelph, CANADA? Canada is a very large, somewhat remote place but it is made up of smaller governmental divisions known as provinces, many of which are known to non-Canucks. Specifying a location as "Canada" narrows it down to anywhere in an area of a little over 9 million square kilometers. So, why not Guelph, Ontario, Canada? Or, conversely, Guelph, North America? Or Guelph, western hemisphere? Further along, we find the Animal Medical Center in New York. Would that be in one of the five boroughs or Massena or Dunkirk?

Chuck, you may have just demonstrated Tyler's argument that people who can't work with computers (or google maps) may end up on the wrong side of 'Average is over'. Besides, surely Watson will be available to answer such questions as 'where are you?'

Evidently, it's the writer that's on the wrong side of "average is over". In communication it's the obligation of the communicator to express intelligible information to the communicatee. It would be easier for him to determine the exact location of Guelph and indicate it in his statement than for every reader to do so. For instance, if a location is given as "Springfield, USA", where is it?

I found the most irritating part of that sentence to be the use of "animal doctors" instead of "veterinarians"

I wonder if Watson can find Waldo.

It is difficult but technically possible to train Watson to find Waldo. However, since there is currently no shortage of six year olds willing to do it for free it wouldn't be an efficient use of resources. Mind you, it was preliminary training for Watson to find cancer in images of biopsy slides it might well be worth it. Six year olds generally lack the patience for that sort of thing.

Whut, no Skynet jokes? :)

Thanks for the write-up, Dr. Cowen. Interesting to see the real-word use cases for IBM's best-in-class system.

It might be fun to see what I can do to optimize the ballistic missile systems now that you mention it.

The medical stuff pleases me. I'm hoping it eventually becomes a good supplement to health care workers in the various retail clinics that are popping up everywhere.

Interesting use of Watson in medical research from Baylor College of Medicine. Not only an intriguing use of Watson for discovery but also suggests a more valuable model for archiving and finding value in the ever expanding db of scientific literature.

Well worth a read.

Comments for this post are closed