Azeem Azhar

really easy and accessible sort of pseudo-authenticity because the tools are getting better and better.

And it took me back to a speech I gave about 15 years ago where I said, look, in the age of AI, and 15 years ago were before AlphaGo and all these other things, right?

Conceptually, making stuff will become perfect because the AI systems, you know, we weren't in the transformer architecture at that time.

will make things to machine calibration.

And in a world where everything becomes perfect, we will start to value the complement of perfection, which is the imperfect, the thing that has some meaning.

And I describe it slightly tongue-in-cheek as the future being artisanal cheese.

You know, we can all go out and get cheese and buy lots of cheese

that's been made in a factory, but there is still some pleasure in artisanal cheese or micro brewing.

And that's what I meant, which is that the machines can produce perfection.

So the thing that's hard or the thing that's human, the thing with texture, with idiosyncrasy, with interiority.

The idea is that authenticity is proof of work.

And I've discovered this because, you know, I'm writing my new book at the moment.

And of course, I'm using the LLM tools to help me with research and with being red teamers on the concepts.

But if you try to get the LLMs to write anything, even a network of LLMs, what you get is incredibly mid-copy.

It doesn't have that interiority.

It doesn't have that idiosyncrasy.

It doesn't have the details that come out from being a human who's lived through something and has experienced things and knows when to break the rules and when not to break the rules.

And there's some empirical evidence around this.

The Arc AGI General Intelligence Benchmark, the Arc AGI-2, showed that pure LLMs score 0% on tasks requiring genuine fluid intelligence.

They're good