Sir Arthur Conan Doyle’s Watson was, in fact, a medical doctor. IBM’s Watson is not a doctor, but tried to play one in real life, and apparently failed. The idea was to use machine learning to crunch huge amounts of medical records and treatment data, and provide recommendations to real doctors treating real patients. And according to this article in Slate, it just didn’t work and the division is being “sold for parts”. I assume this means part of the IBM company legal entity and/or its “intellectual property” being sold, not the actual computer hardware which must be obsolete by now?

Along with the recent Zillow house flipping failure, this seems like another high profile failure for machine learning/AI-based business plans. It might be that the business plans are ahead of the technology, or the technology is ahead of the data (one gets tired of the phrase “garbage in, garbage out”, but it is a real thing – a lot of what is in my medical records is garbage, anyway), or both.