News for hackers. Stuff that matters.

A comment on AI models collapse when trained on recursively generated data. Parent comment

Re:AI models collapse when trained on recursively generated ... by visarga on Wednesday July 24, 2024 @02:16PM (#41060006)
> Meanwhile OpenAI, Anthropics, trains on AI generated data to improve their models, and it works.

They got a secret ace in their pocket - chat logs created with human in the loop. Of course those might still have errors, but much fewer. They can infer from a human response if it was accepted or not.

I think OpenAI generates at least 1B sessions per month and 2 Trillion interactive tokens. Those can go into the LLM again for analysis and synthetic content generation, or for RLHF with the whole conversation as guidance. Having access to the following interactions can shed light on previous answers.

Even more, they can correlate chats across days, presumably humans try out LLM ideas in reality and return for iteration. That way LLMs indirectly get real world grounding.

[ Parent | on HN ]
Re:AI models collapse when trained on recursively generated ... by mrbungie on Wednesday July 24, 2024 @06:31PM (#41062909)
This is likely one of the main reasons why they're offering ChatGPT for free and running ChatGPT Plus at a loss.
[ Parent | on HN ]
Re:AI models collapse when trained on recursively generated ... by apwell23 on Wednesday July 24, 2024 @09:55PM (#41064078)
> main reasons why they're offering ChatGPT for free and running ChatGPT Plus at a loss.

As opposed to what though? Its not like there a huge demand for these apps that they can charge money. They have no option but to give it away for free .

[ Parent | on HN ]
Re:AI models collapse when trained on recursively generated ... by LordDragonfang on Wednesday July 24, 2024 @11:31PM (#41064480)
Are you not aware that half the industry is using it to generate (at least some portion of) their code? And that many are paying for the privilege?
[ Parent | on HN ]
Re:AI models collapse when trained on recursively generated ... by apwell23 on Thursday July 25, 2024 @08:27AM (#41067831)
Yes I pay for it. Not sure what your point is. I didn't say no one in the world pays for it.

Are you not aware of flatlining user growth? or are you under the impression that coders paying for these apps are enough to make them profitable?

https://www.thewrap.com/chatgpt-growth-2024/

[ Parent | on HN ]
Re:AI models collapse when trained on recursively generated ... by astrange on Wednesday July 24, 2024 @07:10PM (#41063168)
They can't directly train on chat transcripts, because they contain private information and other things you don't want appearing in answers. I doubt they even look at them unless you press the thumbs down, in which case they probably use it in some indirect way.

They might try to look for trends or what questions are popular of course.

[ Parent | on HN ]
Re:AI models collapse when trained on recursively generated ... by ipaddr on Wednesday July 24, 2024 @10:35PM (#41064253)
That's exactly what they are doing and what you agreed to ans why some use other models or running something locally.
[ Parent | on HN ]
Re:AI models collapse when trained on recursively generated ... by fantasybuilder on Wednesday July 24, 2024 @11:25PM (#41064450)
Haha, exactly. I switched to local Llama 3 and never looked back.
[ Parent | on HN ]
Re:AI models collapse when trained on recursively generated ... by astrange on Thursday July 25, 2024 @04:00PM (#41072733)
If they were training on other people's chat transcripts, the answers would read like how other people type, instead of telling you to delve mystically into intriguing questions.
[ Parent | on HN ]
Re:AI models collapse when trained on recursively generated ... by ipaddr on Thursday July 25, 2024 @05:12PM (#41073449)
https://help.openai.com/en/articles/5722486-how-your-data-is...

Updated a week ago:

ChatGPT, for instance, improves by further training on the conversations people have with it, unless you opt out.

[ Parent | on HN ]
Re:AI models collapse when trained on recursively generated ... by astrange on Friday July 26, 2024 @01:45PM (#41080601)
Easier to write that than explain their internal processes. But what would be the point? Training on someone asking a question doesn't cause it to learn the correct answer.
[ Parent | on HN ]