News for hackers. Stuff that matters.
Re:The Collapse of GPT by AstroBen on Friday May 16, 2025 @08:02PM (#44010886)
I'd imagine it's really low quality data. Most or all of my conversations with an LLM are questions or telling it to do something, with varying levels of specificity

I'm not sure what they'd get from training on that

Re:The Collapse of GPT by insin on Friday May 16, 2025 @08:08PM (#44010917)
I sometimes wonder if they're vulnerable to a coordinated effort of deliberately upvoting shit assistant turns and praising in the next user turn - how much does that actually contribute to future training, if at all?

I had a very basic React question about useState while porting some vanilla code last week which all models of all stripes I've tried it on have been confidently and completely incorrect about, up to stating the code absolutely will not work, even when I take a turn to assert that I ran it and it does, so there's plenty of shit in there already.

Re:The Collapse of GPT by ted537 on Friday May 16, 2025 @08:30PM (#44011047)
I don't think it would be too hard to scrape useful data out of my LLM convos.

If human response is "That's BS", "fuck off", or something similar, mark as bad assistant message.

If human response is "huh" or "cool", mark as good assistant message.

If on ChatGPT, watch how much scrolling user does. If there's a lot, its somewhat likely that the LLM outputted something useful.

That strategy would have holes of course but as long as its better than guessing something like that would be a useful heuristic.

Re:The Collapse of GPT by londons_explore on Friday May 16, 2025 @11:34PM (#44011845)
This.

Even very weak human signals can be immensely valuable over large enough datasets.

Re:The Collapse of GPT by DeepYogurt on Saturday May 17, 2025 @12:47AM (#44012106)
> If human response is "That's BS", "fuck off", or something similar, mark as bad assistant message.

Marking is not a trivial task though. Use some AI system to mark it and you get a 99.something% filter maybe but whatever that remainder is leaks through. Over time your filter may get worse as a result.

Re:The Collapse of GPT by ehecatl42 on Saturday May 17, 2025 @04:32AM (#44012913)
I'm in the process of messing around with a new distro where things are not quite what I am used to, and the usual suspects have been pretty helpful there... except for when they just make shit up

Grok is the only one that swore back at me. I kinda liked that. The others are way too polite, "Artificial Intelligence? Artificial Canadians, more like", my uni-going kid joked.

Re:The Collapse of GPT by morkalork on Saturday May 17, 2025 @04:13AM (#44012842)
Every time you tell it to do something, it does, and you don't correct it that's a weakly positive signal. If you tell it to do it again with further clarification, that's also a signal. Sometime I feel like I am giving them free work when chatting.. I guess the trade is sort of equitable. Answers in exchange for data..