What did you see back then that convinced you that engineers would actually hand off real work to AI?

Sure.

Yeah, I think coding was one of the applications where very early on it was obvious it was going to be useful.

If you're a software engineer, you're probably one of the first people to hear about the GPT-3 API.

There's such a broad set of things that you need to know when you're a programmer that include how to use Twilio and how to connect to the Twitter API and stuff like that.

that this was just an immediate to answer questions for you.

And in fact, there's a service called Stack Overflow that's effectively like Reddit for programming.

white-collar knowledge worker will be able to do,

A few things there.

Sure.

Yeah.

There's lots to unpack there.

So CodeGen started in late 2022, and at the time, the GPT-3 API had just opened up, so it was sort of the dawn of the modern age of language models.

Everybody was trying to figure out what can you actually do with these things, and very early on, I locked into programming as the ideal application domain.

There's a number of reasons for that.

So how-- talk me through that a little bit, you know, getting that cultural kind of buy-in and then maybe, I guess, from a, a management like perspective, how do you ensure everybody is doing this, if you do, or is it just based on trust that, you know, everybody is doing it?

Oh, you see, I, I don't think I've heard this one, so this is gonna be new news to me.

So this is, this is, this is the, uh, this is the learn feature.

This is a derivative of, of, uh, the AI Assist, which is the, uh, talk to your data feature.

So when we were testing this, this is one we hadn't planned, but it came up in testing and w- this was, like, quite candidly, it's the easiest of the four to do, and people were most excited about it.

And so you don't, you can't predict these things.

And I was just...

You know, I'll tell you, this is real.

Yeah.

there you go you landed the plane beautifully there we go and so cogen yes so from that book so

anyways that sort of kicked me off on getting really interested in this stuff um I went through, I worked at Palantir, I studied AI in college.

I did a bunch of machine learning stuff, did a decade's worth of startups, give or take.

And around 2020 is when, so I've always kept up with the state of the art of AI research.

It's just sort of something that's very interesting to me.

2020 is when the GBT-3 paper came out, which was essentially just scaling up language models and ended up having these incredible properties of basically few-shot learning.

So you could give it pretty much any problem that we were dealing with in NLP at the time.

It's the world model.