I am not surprised at all.
Background: I have spent my whole career in the applications of AI/ML/Data Science to healthcare and the life sciences. In diagnostics, devices, drug development.
Disclaimer, some of this going to be repetitive from my own comment history, such as from back in June when there was a story about Apple struggling in Healthcare:
https://slashdot.org/comments....
The folks at Watson never really understood healthcare. They figured hey we did a great job with Jeopardy, that was AI, now we can do AI in healthcare. But first of all, Jeopardy was a big NLP knowledge graph. Sure that is AI, but the AI for healthcare applications is much more. And even then, you can not just throw âoeAIâ at healthcare, you need to understand all the different subproblems underneath.
Are you talking about drug discovery, medical claims and costs, diagnostics, treatment, devices, population health, genetics, providers, hospital networks, insurers, pharma, etc? What type of data are you dealing with: genetic, omics, imaging (digital pathology, radiology, mass spec, etc), RWE/claims, devices/streaming, behavioral, population health, chemical/molecular/drug entity, etc? And much more. Each has its own nuances, issues, problems, algorithms, approaches.
Slashdot user tomhath previously said it was a
solution looking for a problem, pretty good at parsing English
but not more than that. User technomom has said
It is a brand, nothing more. The original code itself was purpose written specifically to play Jeopardy. There was nothing in it appropriate for medical or other applications.
https://slashdot.org/comments....
https://slashdot.org/comments....
Over the course of my career, I have had multiple different interactions with the folks at Watson Health. It is a huge org, I am sure some people there got it, but I never met anyone who really did. I visited the fancy showcase/partnership development center they have in Boston with all the fancy digital and AR displays.
Looked AWESOME. Lots of money sunk in to impress the hell out of how shiny and cool everything is. It is IBM after all. But a very Potemkin construct.
Once you actually start talking to the people about what they had, how it was developed, how it was validated, why could it be trusted, what circumstances was it strong/weak in, future roadmap, etc. Anything about the real meat and bones. The conversation collapsed. Non technical talking points were regurgitated (we are IBM, this is cool, we know AI, we are working with MD Andersen). It was obvious that the REAL issues of AI in healthcare were not dealt with, it was all quick and dirty pilots, proof of concepts, things slapped together.
At conferences and various industry fora (in the pre-Covid times) I would meet lots of people working in Watson Health. They either did not last long there, or very flashy visionaries talking about broad generalities. There did not seem to be anyone there who could actually get stuff done on the ground. If they were, they were very frustrated I am sure. IBM threw money, charlatans who make careers off big ideas worked there for a bit, but in reality nothing bore fruit. Where were all the real doers?
And you know what, healthcare is HARD. Complex. You would never set things up this way from scratch, but there are reasons why things are the way they are. Scaling things up and dealing with humans, messy, sick, biology with all itâ(TM)s various complications and issues. The recent experiences with Covid should just highlight how messy it all is.
People will do one proof of concept or pilot. In one sub sub sub disease, on a small dataset, from a particular set of dates. Well medical practice changes, is that dataset from 3 years ago reflective of medicine today? That dataset was for men 20-40, does it work for women over the age of 50? It had samples from the Mayo clinic, is that how samples and patients are treated at the ER in Podunk, Middle of Nowhere? OK it worked for that one disease, how about for the other 5000? And on and on.
IBM has struggled in its main tech areas. And even people who do much better in tech (Google, Apple) are flopping in healthcare.
Good rule of thumb:
1. It is more complicated than you think
2. If you think it is very complicated, you are wrong. It is much, much worse.
I could keep going, but this is already a pretty long post. I guess the TLDR is that IBM thought we have a great âoeAIâ brand, letâ(TM)s throw money at healthcare. And never really cared to figure out healthcare and do it right.