Peter H Diamandis doesn't understand anything about AI or medicine beyond the most superficial stuff. We have had clinical decision support systems for decades that produce good accuracy for diagnosis. This doesn't particularly require ML or LLMs or neural networks or any of the recent AI hype technologies. Older, simpler statistical techniques work just fine. However, these existing CDS have never become part of the standard of care because there isn't much evidence that they improve patient outcomes. Most diagnoses are fairly straightforward and don't require a CDS to get it right. In order for a CDS to really shine in the rare "zebra" diagnoses it takes a lot of extra work for clinicians to do data entry, and they don't have time for that.
I think he might. He has an MD attached to his name after all. But I'm pretty sure the AI he's specifically referring to is something in relation to large language models, or he's simply just unaware that the technology you mentioned is being used at present.
