[343rwerfd]

nor OpenAI or any of the prompt-based AI companies actually "need" the reveneu from the services they sell, the whole point of having a public (free or not), prompt facing the entire planet is just having live humans doing RLHF 24x7x365, that information, that dataset is more valuable than any symbolic amount of money anybody is willing to pay for any GPT or clone suscription.

Anybody noticed already that any current or near future reveneu won't make a dench in the actual costs of running giant models, anyway, the models keep chugging along just fine. And the ("free") money keeps flowing in.

[343rwerfd]

Moroever, those millions of non-paying clients prompting the models are 24x7x265 working with 100% real world problems are inputing the models with valuable prompts, generating valuable content (originated in real situations, actually distilled by unvaluable billions of human sensory input).

That content can be and is used to train models, effectively cancelling the "data-wall", bit by bit, all the time.

[jncfhnb]

> That content can be and is used to train models, effectively cancelling the "data-wall", bit by bit, all the time.

It really cannot.

[CoastalCoder]

I'm not familiar with this topic.

Could you elaborate?

[jncfhnb]

There’s so just much missing information or changing contexts. Forget any specific model. You will not become particularly good at any domain specific task by reading prompts of people asking questions. That data sucks.