The Wall Street Journal(USA) August 11, 2018
Big Blue spent billions on its Watson artificial intelligence product with a focus on cancer care. Sometimes, it didn’t say more than doctors already knew. In other cases, Watson was tripped up by a lack of data in rare or recurring cancers and rapidly evolving treatments.
This comes as no surprise to me. I have spent my entire career as an engineering mathematician looking at cause and effect and how to predict things that can go wrong ahead of them failing. First under Admiral Rickover in the nuclear navy working on avoiding a repeat of the Thresher disaster where the submarine and 129 crew were lost during sea trials, then for the New York City Hospital Association predicting for staffing purposes who will show up in the emergency room on a hot summer evening, then for Mechanical Technologies advance equipment designers predicting bearing failure to enable predictive maintenance in offshore high-speed rotating equipment, the story is always the same.
You have to start with “training data” to drive the analysis. It is not about some big black box, or in this case a Big Blue box that somehow works a miracle, despite what the IBM ads seem to indicate in the commercials we often see on TV. That is marketing hype and the realities do not live up to that hype.
We at Apogee have a long history of training data from our staff’s collective careers studying first-hand energy use in homes and businesses. I have personally field audited almost every type of large US industrial plant. We know what is hidden in bills and meter data because we have seen it operate and then seen it show up in electrical and fuel use.
That is why our analytical engine was shown by EPA, DOE, and LBL as twice as accurate as any other in the world in their year-long testing. The idea that someone can simply throw data into a box, stir it around and get useful results is nonsense.
Thank you IBM for finally admitting it.