News for hackers. Stuff that matters.
Re:The Collapse of GPT by keybored on Saturday May 17, 2025 @01:50PM (#44015844)
It’s the same argument we see again and again. Someone might say that we need “human cultural artifacts”. Then someone says, “but what is human cultural artifacts”? Then they follow up with how rubbing neurons together in response to stimuli is in principle as mechanistic as whatever language models do. From there they lean on incredulity: well I don’t know anything about human beings other than reductionist tropes, but I sure see no reason to make any distinctions between silicon and carbon.
Re:The Collapse of GPT by js8 on Saturday May 17, 2025 @02:35PM (#44016115)
It's not the same argument because I don't make any assumptions about how LLMs work. All I am saying that people have been able to keep reality in check in the presence of cultural artifacts, and will continue to do so even if such artifacts are produced by AI. Because what makes these artifacts interesting to humans is their relation to the (truth of the) real world, and that's regardless of who or what produces them.
Re:The Collapse of GPT by keybored on Monday May 19, 2025 @03:39PM (#44033942)
My mistake. It’s just pure incredulity.

- Humans have done this for centuries

- These are all cultural artifacts

- ?

- It’s all the same to me

People have claimed—and this is a widespread theory—that a lot of LLM brilliance comes from datamining creative thoughts/output from humans and that the brilliant insights go down when there isn’t much in the way of such material. Further they claim that the eventual convergence towards LLM-only content (once humans presumably give up) will not generate the same quality of output. In fact it will deterioate.

Maybe someone would like to contest that. But that should be done directly. Instead of making pedestrian statements like:

> Humans have been trained on "human-generated data" (cultural artifacts) for centuries, and quality is not down.

Which is borderline just a rhetorical gotcha.

Re:The Collapse of GPT by js8 on Tuesday May 20, 2025 @12:07PM (#44043174)
> Maybe someone would like to contest that. But that should be done directly. Instead of making pedestrian statements

I am not sure what you want from me here.

Yes, I do contest that "that a lot of LLM brilliance comes from datamining creative thoughts/output from humans and that the brilliant insights go down when there isn’t much in the way of such material".

Because I think, the same is the case with humans. Most of the cultural artifacts we produce is crap, a bad copy of natural original. And the "brilliant insights of humans" are achievable by models running on higher temperature.

I think the proponents of the theory need to explain by which mechanism the actual loss of information supposedly occurs (on the probability distribution of possible LLM outputs). Is it averaging? Added randomness? Preferential skew? To me, it is rather vague, to the point I don't see how it's different from what humans have done for centuries.

Or the opposite, show how those "brilliant insights" from humans manage to survive in the sea of cultural crap otherwise produced in human culture. Perhaps a specific example would help.

Re:The Collapse of GPT by keybored on Wednesday May 21, 2025 @11:07AM (#44052289)
> I think the proponents of the theory need to explain by which mechanism the actual loss of information supposedly occurs (on the probability distribution of possible LLM outputs). Is it averaging? Added randomness? Preferential skew? To me, it is rather vague, to the point I don't see how it's different from what humans have done for centuries.

This is the pure incredulity that I was talking about.

Re:The Collapse of GPT by js8 on Thursday May 22, 2025 @03:58AM (#44059776)
No, it's skepticism. I think proponents of that hypothesis should come up with some testable prediction, how is that supposed "loss of quality" affecting the distribution of possible LLM outputs. Then we can have debate on it's effects and relevance to existing human cultural output. (But I already ran over various options, and the above is the conclusion I came to.)