Ed Elson

You're describing...

that the model gets better when it's subjected to, or when it's fed large amounts of data and also like diverse forms of data.

And originally we were kind of just limited to the web, but the web isn't all of life.

There's more beyond the web that these models could be trained on.

And so too, you could say the same thing about these chats.

I'm wondering if there's other forms of data that you think will be

prevalent for model training in the future.

You know,

things in the physical world.

I mean, typing words onto a keyboard and seeing words on a screen isn't everything.

But to AI right now, it seems to be close to everything.

So is there a way, are there other forms of data that you think in the future AI will be fed and therefore that would sort of take us, I guess, on the path to AGI?

Do you think that AI, the people building AI, the leaders of the AI industry, Sam Altman probably being the high priest right now at least, do you think that there's not enough appreciation of that?

Do you think that people are too obsessed with, we need AGI, we need human-like intelligence?

I just look at the contract between Microsoft and OpenAI, which basically one of the stipulations in the contract is the

The terms will change once we achieve AGI.

I mean, there are many questions like, what does that even mean?

But the fact that AGI is sort of the benchmark for everyone, and I'm even asking you, like, how do we get to it?

Do you think there's too much obsession with this concept of AGI in the AI industry right now?

Yeah.