Challenges and Lessons from IBM Watson's Early Partnerships and Data Acquisition (Full Transcript)

Speaker 1: big challenge. But my personal story here is that when we were a tiny little startup, we were actually an IBM Watson partner. And we were like, I'm serious, it was before their partner program was even in place. And we were like, wow, this is amazing, you know. And we couldn't, once we figured it out, we figured it out. It makes, this makes also makes perfect sense is that IBM doesn't actually have any data. So they have to, they have to partner with anyone possible that whether that's a health system or another vendor to get the data. And they are, I want to say, like, I think they're actually a little, they're alone in this in the industry. Like every other technology company has some place that they have some data that they can analyze and then compare things to, you know, like Google's got all the search data and, you know, was doing the work with FluFinder for a while. But I think the biggest challenge with IBM was that they had nothing to start with. And if you don't have data, you can't train your algorithms. And so no matter how smart your machine learning is, if it's not being trained on anything, it's not going to get any smarter. So it's an interest, it's, you know, it's great moonshot, but I think fell down in the execution and the key execution of like, where are we going

Speaker 2: to get data from? Yeah, it's, well, and they went out and bought it, right? So they went out, they bought Explorys. And Explorys had data from a fair number. I mean, Explorys was pretty big at the time they bought them, a fair number of health systems. And we were one of those health systems in Southern California. We were also an investor in Explorys. So we were happy to see them, you know, with an exit. It's not in the unicorn exit status, but it was a pretty good exit at the time. Yeah, you were talking about this. Let's see, 2013, they started with the MD Anderson pilot to mine for insights from health systems, vast groves of research and patient data, and develop new NLP power decision support tools. By 2018, the two organizations had fallen out with MBA Anderson pulling the plug on the project after spending more than $60 million, following multiple examples of unsafe and incorrect treatment recommendations. Yeah, and I remember when that happened, because they came in, they had the big booth at HIMSS. They were talking about all the things that we're going to do. I even went to a session. I don't remember if this was at HIMSS or CHIME, where the IBM people were talking about what they could do for us. And essentially what they mapped out was a normalization of our, sort of a data governance slash normalization of our data to make it ready for AI. And I thought, that's interesting. And by the way, the project they mapped out for me was probably a $60 million project just to clean up the data. And I was like, wow, that's amazing. And by the way, I still think AI and machine learning on the clinical side is still highly risky.