Azeem Azhar

And if I then write about something that I think he's done that's really impressive, SpaceX, the inbox fills up and it's not the same people and they're all cross.

And so there are these hot buttons where you can't at the moment find a way of getting people to sit around the table and say, listen, we can accept that this person does something impressive here and they do something difficult here.

And that's all within the capability of a single person personality character.

And I just wonder, you know, are there historical precedents for us going through these moments?

And what are the things that we can learn that help us come back together, bring it back to, you know, to a sense of common understanding while still being able to disagree quite civilly?

I don't write about ketamine that much.

I mean, we've written about psilocybin.

We've written about GLP-1s.

But I mean, I love this idea.

I mean, we do use AI tools in that way.

I mean, we use AI tools to challenge perspectives, to find weaknesses in arguments.

Well, on Saturday, we've got a really interesting one that I think plays into the limits of AI.

So earlier this week, I wrote something about Donald Trump and Senator McCormick putting something together to activate tens of billions of dollars in an industrial policy to generate more electricity.

US turning to industrial policy.

And we describe McCormick as a Senate hopeful.

He is a sitting senator since January and human fact checkers are there and there are LLM fact checkers.

And the LLM fact checker didn't spot it because it thought that the big piece was, you know, the $80 billion, which is true in some sense.

One of our readers pointed this out.

As we started to diagnose what went wrong, I put more and more sophisticated prompts into different LLMs, grok, perplexity, chat GPTs 03.

And after six iterations, it was still not spotting the mistake.