Sam Altman: Getting Fired (and Re-Hired) by OpenAI, Agents, AI Copyright issues

health care, I believe, should be pretty transformed by this. But the thing I'm personally most excited about is the sort of doing faster and better scientific discovery.

And they've basically kept all the IP for AlphaFold 3 in Isomorphic. So Google is going to monetize the heck out of this capability.

It seems like a car is the only thing people can think of that would move the needle in terms of earnings.

Apple hasn't bought anything more than $50 or $100 million. And so the idea that all of a sudden they come out of the blue and buy a $10, $20 billion company, I think is just totally doesn't stand logic.

if you're running out of in-house innovation and you can't do M&A, then your options are kind of limited. I mean, I do think that the fact that the big news out of Apple is the iPads getting thinner does represent kind of the end of the road in terms of innovation.

It's hard for me to imagine that we just go to a world totally where you say, like, hey, chat GBT, order me sushi, and it says, okay, do you want it from this restaurant? What time? Whatever. I think visual user interfaces are super good for a lot of things.

Well, the quality is not good, but when the quality is good enough, you'll actually prefer it just because it's just lighter weight. You don't have to take your phone out. You don't have to search for your app and press it.

Basically, the California legislation that's proposed and some of the federal legislation that's been proposed basically requires the government to audit a model, to audit software, to audit and review the parameters and the weightings of the model. Then you need their check mark in order to deploy it for commercial or public use.

the reason I have pushed for an agency-based approach for the big picture stuff and not write it in law, is I don't, in 12 months, it will all be written wrong

The things like device companies or if we were doing some chipfab company, those are not Sam projects. OpenAI would get that equity.

All those projects, he said, are part of OpenAI. That's something people didn't know before this and a lot of confusion there.

So I think there's an opt in, opt out in that case, first of all, and then there's an economic model.

What's the difference here and why are you making the case that perhaps artists should be uniquely paid? This is not a sampling situation. The AI is not outputting and it's not storing in the model the actual original song. It's learning structure.

The music industry is going to consider this opportunity to make Taylor Swift songs their opportunity. It's part of the four part fair use test is, you know, these who gets to capitalize on new innovations for existing art.

I think maybe the two big vectors, Sam, that people always talk about is that underlying cost and sort of the latency that's kind of rate-limited, a killer app.

We want to like cut the latency super dramatically. We want to cut the cost really, really dramatically.

I'm not sure there's much more to add on the HubSpot acquisition rumors. They are still just rumors, and I think we covered the topic a couple of weeks ago.

I think the conversation has been historically very caught up on training data, but it will increasingly become more about what happens at inference time.

I think the world needs a lot more AI infrastructure, a lot more than it's currently planning to build and with a different cost structure.

Same with how I'm, like, more excited about humanoid robots than sort of robots of, like, very other shapes. The world is very much designed for humans and I think we should absolutely keep it that way.

you had OpenAI say we're going to basically put the whole company and work together to make GPT-4. And that was unimaginable for how to run an AI research lab. But it is, I think, what works.

I think that these guys are going to be one of the four major companies that matter in this whole space. I think that that's clear. I think what's still unclear is where is the economics going to be?

the accuracy or the value of these models will probably shift to these proprietary sources of training data that you could get that others can't

I definitely don't think it will be an arms race for data because when the models get smart enough at some point, it shouldn't be about more data, at least not for training.

when Llama 3 got released, I think the big takeaway for a lot of people was, oh, wow, they've like caught up to GPT-4. I don't think it's equal in all dimensions, but it's like pretty close or pretty in the ballpark.

What we're trying to do is not make the sort of smartest set of weights that we can. What we're trying to make is like this useful intelligence layer for people to use. And a model is part of that. I think we will stay pretty far ahead of, I hope, we'll stay pretty far ahead of the rest of the world on that.

these models will roughly all be the same, but there's going to be a lot of scaffolding around these models that actually allow you to build these apps. So in many ways, that is like the open source movement. So even if the model itself is never open source, it doesn't much matter because you have to pay for the infrastructure.