There was never anything open about open ai. If there were I should have access to their training data, training infra setup and weights.

The only thing that changed is the reason why the unwashed masses aren't allowed to see the secret sauce: from alignment to profit.

A plague on both their houses.

They don't publish papers now, they actually published papers and code before.

No doubt OpenAI was never a glass house... but it seems extremely disingenuous to say their behavior hasn't changed.

What as "open" about it before that?

The first word in their company name.

Isn't Ilya even more against opening up models? OpenAI is more open in one way - it's easier to get API access (compared to say Anthropic)

What was "open" before ChatGPT?

In terms of the LLM's, it was abandoned after GPT-2 when they realised the dangers of what was coming with GPT 3/3.5. Better to paywall access to and monitor it than open-source it and let it loose on the world.

ie. the original mission was never viable long-term.