IBM Carves Out Watson Business Headquartered In The Big Apple

January 13, 2014 Timothy Prickett Morgan

Big Blue wants its analytics business to grow to $20 billion by the end of 2015, and it wants its Watson question-answer machine to do more to help the company make those numbers. And so, last week in New York, Big Blue created a new business unit, called the Watson Group, to further commercialize the technologies that are at the heart of the Jeopardy! playing system.

IBM has been working with Memorial Sloan Kettering Cancer Center, WellPoint, MD Anderson Cancer Center, Cleveland Clinic, and Elsevier on various uses of the Watson expert system in the medical field. Sloan Kettering is using it to give doctors advice on cancer treatments, and the Cleveland Clinic is using it as a teaching tool for doctors in residency. But there are other uses. Fluid, a hip young retail software maker, is using Watson technology to create a personal shopping assistant and has prototyped it at The North Face, and Travelocity and Kayak founder Terry Jones, who was formerly the IT director at American Airlines’ Sabre unit when Travelocity was launched, hinted at how a Watson front end for travel and hotel sites could radically change the way we choose and book vacations. Jones was a travel agent four decades ago, and he conceded that when he books his vacations, he still uses a travel agent for their expertise.

The complexity of choices and the depth and breadth of data in healthcare, finance, manufacturing, retail, and travel are, in essence, forcing humanity to give up on trying to gather up scads of data and learn everything about a topic to accomplish a task or trying to reach out to experts who may or may not be available when we need them. So, the theory goes, we instead put the thinking of experts into systems like the Watson and have it make connections between old and new data and give us probabilistic–rather than deterministic–answers to our questions. How to treat this cancer or where to go on vacation, whatever.

That, in a nutshell, is what the Watson Group is about delivering, and IBM is shelling out $1 billion over the next few years to turn the Watson software into a set of products that others can use to create applications. Think of it as WebSphere, only with brains. IBM wants Watson in the middle of everything.

“It is not a super search engine,” explained Ginni Rometty, IBM’s chief executive officer and chairman, at the launch event in New York. “It can find a needle in the haystack, but it understands the haystack.”

This is a lot harder than it sounds, and that is why it has taken two years to get Watson doing very specific–and limited, however useful–tasks in the medical field. According to a report in the Wall Street Journal that came out just ahead of the launch of the Watson Group, late last year IBM’s top brass were not happy with how quickly Watson was being commercialized. The goal, according to a transcript of IBM executives discussing Watson, was to have Watson generate $1 billion in revenues by 2018, but total sales of Watson products (presumably through its cloud service that went into beta last year as well as software licensing and related hardware and services revenues) was under $100 million.

Watson Group SVP Michael Rhodin

Since Watson won the Jeopardy! showdown with humanity two years ago, IBM has been clear that it believed that the third era of computing–cognitive computing, where machines learn from data and context and are not, strictly speaking, programmed using a high-level language–was upon us, following the tabulating and programming eras that Big Blue was a dominant factor in. The punch card tabulator and the mainframe and PC are all inventions that IBM either created or commercialized with businesses and profited from handsomely.

The new Watson Group will use $100 million of its budget to invest in startups that want to create applications based on the Watson question-answer system, and the Watson cloud that IBM put into beta last November already has 890 different developers or companies interested in using the tool. It looks like IBM will be trying to sell Watson as a service more than a product and will also be creating interfaces between its DB2 databases, its Hadoop distribution, and its Cognos and SPSS analytics tools.

Michael Rhodin, who is the senior vice president in charge of the new Watson Group, also said that IBM was working on extending Watson so it can hear and so it can process both pictures and video because these are also part of the ways that humans interact with each other. IBM has created a number of Watson services to sorting through data as well as for automagically loading datasets, finding the interesting bits based on some questions, and displaying that data. This last one is called Watson Analytics, which was developed under the code-name “Neo,” and it looks like the most interesting one to my mind. It goes into beta testing this week.

As usual, I have some concerns with any new technology, especially one so broad and deep as Watson. Automation is ever the way with humanity, starting with the establishment of agriculture and animal domestication somewhere after the last Ice Age ended 13,000 years ago. Someone figured out how to bring order to the tasks of hunting and gathering, and they reaped the benefits by not needing to have to hunt or gather any more. Progressive levels of automation have been improving our lives while eliminating our jobs for centuries, and at a seemingly ever-faster pace. IBM has been very, very careful about pitching Watson as a collaboration tool, but it is not hard to imagine that a Watson system would be better at answering queries at a call center for a lot of calls, or worse still, that there will be two tiers of most services. Those who get the Watson-based service directly and those who pay more and get a human expert that has access to a Watson expert system.

I also think that IBM will have to be very careful to not try to charge monopoly pricing on Watson’s software, whether it is licensed or sold as a service, because that will only encourage others to compete. Google, for one, has the smarts to create a Watson alternative. And customers relying on Watson services will be antsy about lock in, and that will absolutely compel competition. But, as VMware has demonstrated so well with its X86 server virtualization, you can sure get away with charging a premium for a premium product for quite a number of years until the competition catches up. And I expect IBM to do this.

What this has to do with the IBM i platform is not precisely clear. The original Watson system that played Jeopardy! was a cluster of 90 Power 750 machines with 2,880 cores and 16 TB of main memory. IBM said in its announcements that the latest iteration of Watson could fit in three rack servers the size of a pizza box and offer 2,400 percent higher performance. IBM has not crammed that much more performance into a server, so it has clearly unthrottled and tuned Watson. There is no reason I can think of why a baby implementation of Watson could not run on a set of Power Systems partitions, but the speed of its querying and answering is based on having a fair amount of oomph. IBM seems more interested in hosting Watson instances on its SoftLayer cloud, and has promised to do so.