[regularfry]

> No, he says it because its true and concerning.

Both can be true. It is extremely convenient to someone who already has an asset if the nature of that asset means they can make a convincing argument that they should be granted a monopoly.

> LLMs seem like a dead end.

In support of your argument, bear in mind that he's making his argument with knowledge of what un-nerfed LLMs at GPT-4 level are capable of.

[ben_w]

> It is extremely convenient to someone who already has an asset if the nature of that asset means they can make a convincing argument that they should be granted a monopoly.

While this is absolutely true, it's extremely unlikely that a de jure monopoly would end up at OpenAI's feet rather than any of the FAANGs'. Even in just the USA, and the rest of the world has very different attitudes to risks, freedoms, and data processing.

Not that this proves the opposite — there's enough recent examples of smart people doing dumb things, and even without that the possibility of money can inspire foolishness in most of us.

[regularfry]

> While this is absolutely true, it's extremely unlikely that a de jure monopoly would end up at OpenAI's feet rather than any of the FAANGs'

Possibly. The Microsoft tie-up complicates things a bit from that point of view. It wouldn't shock me if we were all using Azure GPT-5 in a few years' time.

[ben_w]

It's possible, I don't put much weight on it given all the anti-trust actions past and present, but it's possible.