Back to episode — Episode 3065 CWSA 01/07/26
Context —
ng to escape the US Navy. And so they sent out a sub to maybe protect it. But I believe that as of just a few minutes ago, the US forces actually took it. So they boarded it and they took it before Russia could get any serious navy presence there. But obviously we had to hurry because we didn't want to do it when there was a Russian submarine ten feet away. So another success for the US military.…
← Previous segment →s more fake because it hallucinates. Will we ever get to the point where the AI can fact-check the fake news?
Well, it can definitely fact-check the fake news when it's from independent publishers. I think it was this morning and also yesterday I saw a story that was being promoted by some random account and I read the story. I was like, "Wow, that's blowing my mind if that's real." But then I caught myself. I said, "How would I know if it's real?" So I went to Grok and said, "Is this real?" And Grok said, with no uncertain terms, it was not real.
Now, I tend to be influenced by the last thing I hear. So I immediately ignored that post and I've now seen it pop up as being real. So I think Grok got that one. But how could I know that Grok will get the next one? I mean, it hallucinates all kinds of things about me. How would you know? If you were trying to check a story about me and you went to AI, what are the odds it would be accurate? Not really high. And that's with Grok being the best one in terms of accuracy, I think.
According to the New York Post, federal prosecutors are going to be coming hard for California fraudsters. Something about the homeless services being defrauded now. I don't know. Do you expect to see lots of arrests about California fraud? I think Elon Musk was saying that the fraud in California would end up being way worse than the frauds in Minnesota. That doesn't feel good, but it seems necessary. So we'd like to see more prosecutions and we'd like to see them act faster because whatever we do, it always seems like they could have been a lot faster.
Meanwhile, the Washington Examiner is reporting that Utah will be the first of its kind partnership with the medical community that allows patients with long-term conditions to refill prescriptions using AI. Now, how do you make that work if AI hallucinates? I'm not saying they won't, because if you narrow the domain, AI can do well. And this would be a case of narrowing the domain. And it's only limited to people who have been getting the same prescriptions for a long time. Sounds like a good idea.
And then the AI could quickly check to see if there are other medications that changed since they got a prescription. They'd be able to have the AI tell them if anything's now incompatible. B
Context —
ut I think they have that already, don't they? Doesn't your health care provider have some kind of a program that they've had for a while that would tell them if they prescribed one thing it might conflict with the other thing? Because I've had them tell me that by looking at a screen, unless it's just that doctors are taught what doesn't go with what and that's all they need. I doubt it though.…
Next segment → →