Tyler Cowen

That's good and it's important, but I think it's also of limited value because the more we learn how to control AI systems directly, the bad guys will have similar lessons and they will use alignment possibly to make their AIs bad and worse and that it obeys them.

So yeah, I'd rather the good guys make progress on what they're trying to do, but don't think it's going to solve the problem.

It creates new problems as well.

I'm writing a book right now.

I suspect it will be my last.

That book, its title is Mentors.

It's about how to mentor individuals and what are the social sciences know about mentoring.

My view is that even if the AI could write the book better than I can, that people actually want to read a book like that from a human.

I could be wrong, but I think we should in the future restrict ourselves to books that are better by a human.

I will write every day for the rest of my life, but I'm not sure that books make sense at the current moment.

Well, I just submitted a piece.

It's a defense of elitism.

So the problem with our elites is that they have not been elitist enough and have not adhered strictly enough to the scientific method.

So it's a very simple point.

I think to you it would be pretty obvious, but it needs to be said.

It's not out there enough in the debate.

But yes, sometimes the elites have truly and badly let us down.

But the answer is not to reject elitism per se, but to impose higher elitist standards on our sometimes supposed elites.

So that's the piece I just sent in.

It's coming out soon.