Facing Jeopardy: Which company is finding problems in identifying the right homes for its Watson?
At a recent event, IBM’s Watson was an area for discussion. Watson sprung to fame in the US by beating two humans in a TV games show, Jeopardy! This led to a lot of coverage for IBM—but also to a lot of problems.
Firstly, those that know of the win see Watson as akin to when IBM’s Deep Blue computer first beat a human chess master—it’s a great idea, but it is only a game, after all. Bridging the gap between what Watson did on TV and what it could do for an organisation is proving a little difficult.
Secondly, outside of the US, Jeopardy! is not so well known, and IBM’s marketing around it has often led to a need to explain the programme before moving on to what Watson could possibly do in a real world scenario. The feelings are that using Watson in, for example, Mastermind or University Challenge would run up against the same issue as seen in the US—a great games machine, but so what? On top of that, would IBM then have to enter similar competitive programmes in Germany, France and everywhere else? A knotty problem.
So, just what could Watson really be used for? It seems that many of the discussions IBM gets into with prospects boil down to a perception that it is a pumped-up search engine—which misses the point. Watson is the current exemplar of how to deal with real big data issues—it works against documents as well as formal data sets; it is excellent at picking out key aspects of information thrown at it; and it is getting better at assigning probabilities as to how accurate its results are against how the user is feeding it the questions and supporting data it requires.
Watson really is a fast working, highly knowledgeable assistant to humans who need to work against massive and complex data sets in order to get to a best response. Sounds a bit woolly? Unfortunately, it’s a bit difficult to tie things down much more than that.
However, Watson is moving out from the Jeopardy! cloud in some areas. For example, it is being used in healthcare—from helping to reach a more accurate diagnosis for a patient to helping to calculate the correct approach to continuing care for a health insurance company. The idea is NOT to replace the health professional, but to provide a dispassionate assistant that can use real-time information analysis to drill down and provide a range of options back to the professional.
This requires a degree of focus from Watson, and it is showing its best capabilities where its assistance can be narrowed down considerably. At the moment, Watson is not there to aid a general practitioner: it is there to help—for example, an oncologist can use Watson to assist in not only helping to decide exactly what cancer a person has, but also to come up with advice as to which treatment is most likely to have the best outcome. Even here, Watson works best if it can be further constrained—for example, to concentrate on leukaemia or lymph node cancers. The greater the focus, the more accurate the results.
Does this make Watson too specific for general use elsewhere? Certainly, as an on-premise solution, it is only for very large organisations with a highly specific need. Pharmaceuticals with research needs in a specific area where external information from other sources is rapidly changing is one possible target. Organisations with large patent libraries where "prior art" is important or searching out possible beneficial overlaps is another.
But how about WaaS (Watson as a service)? Is Watson’s specificity a major block to a capability here? Look at the legal sector—many organisations are small and could not warrant investing in an on-premise Watson implementation, yet having a capability to use a Watson approach in, for example, case law where an assistant that can sift through legal precedent and other information sources and advise on approach would provide immense help to paralegals, lawyers and barristers (and their equivalents worldwide).
There is great deal of promise from Watson—and, by its very nature, anything that is done in the real world at this stage is not going to be wasted. As Watson improves, it can take the silos of information from early-stage Watson implementations as feeds, using a federated approach to build up a networked 'Super Watson'.
IBM may still be some way away from what the original researches set as a vision—to replicate a computer system as seen in Star Trek. However, the steps along that journey are well under way, and the healthcare examples of real world Watson usage are already showing the strength of the system. Is this Watson’s only spiritual home? It shouldn’t be—but IBM has to be able to more succinctly and effectively message how Watson is different to search engines, business intelligence and business analytics to be able to get its point across.