Tech Thread

[MindzEye]

You're right, I guess what I meant was that his foundation would have been up against the slumbering mass of human systems constantly in their attempts to deploy humanitarian aid. Witness the slow pace of change and cultural and political minefield. Even in the abstract I thought he'd understand this better.

Every time I read a piece where someone is talking about their unscripted, non business related interactions with a billionaire, I come away realizing more and more how they operate like they're basically another species of humans. None of them started out like that either.

Here's a recentish one that I read shortly after it came out: https://www.theatlantic.com/magazine/2026/05/billionaire-consequence-free-reality/686588/ (no paywall https://archive.is/20261001113707/h.../billionaire-consequence-free-reality/686588/ )

Shit just stops meaning anything to them after a while. Gates likely doesn't know shit about his foundation aside from how much money goes to it, that his name is on everything and that it's part of his 'legacy' in that he wants people to remember him well and not just for being a rich douche.

In the interview for example Gates feels in the near term almost all medical work will be done by machines, at least from a diagnostic / decision making perspective. And I mean in an idealized frictionless world I could image how machines would of course be superior in so many way, but I just can't see a path there that doesn't take generations, and as you say doesn't in the end open up new opportunities for human's in the meantime.

and all of that is largely because he's so disconnected from the ground level reality of the subjects he's talking about. As a general rule, the returns on AI in the medical space seems to be that it's closing a few diagnostic gaps, but that it can't do either of them alone without significant human oversight. It's great at diagnosing weird edge cases that Doctor's don't consider because of their relative rarity, it's great at early detection, and it's great at identifying diagnostic errors made by human doctors. But it absolutely requires a human in the decision making loop because it makes a lot of mistakes. The general takeaway is that using them together is superior, by a lot, than using either separately. Now, I know that the AI pumper response to this is "the technology is going to get better" and while that's probably true to a degree, we're definitely in a world where the frontier models are very small improvements on the last with billions of dollars thrown at that slight improvement to make it possible. That's not really a viable business model going forward. There's no ROI on a 2% improvement in coding efficiency at that cost of 10 billion in R&D expense.