Hacker News Reader

trivial to replace

And yet no one has been able to do that since gpt4 was released.

That's only because the key players have no reason to compete.

They don't want to run a developer/enterprise ChatGPT platform.

Google cares about Search, Apple about Siri, Meta about VR/Ads. But those three are interesting heavily in their own LLMs which at some point may better OpenAI.

It's not trivial given current supply bottlenecks, not to mention research expertise.

I don't feel like compute for pretraining the model was a huge constraint?

The supply bottlenecks have been around commercializing the ChatGPT product at scale.

But pretraining the underlying model I don't think was on the same order of magnitude, right?

The control of the supply si with Microsoft, who are likely falling on Sam’s side here.

First mover advantage and Microsoft integration is nothing to sneeze at.

For sure.

But if Altman has a new venture that takes first mover advantage on a whole different playing field MS could easily get left in the dust.