Holy shit are you telling me…
Garbage In…
= Garbage Out?
AI models collapse when trained on recursively generated data
Submitted 3 months ago by floofloof@lemmy.ca to technology@lemmy.world
https://www.nature.com/articles/s41586-024-07566-y
Comments
sp3tr4l@lemmy.zip 3 months ago
FaceDeer@fedia.io 3 months ago
You realize that those "billions of dollars" have actually resulted in a solution to this? "Model collapse" has been known about for a long time and further research figured out how to avoid it. Modern LLMs actually turn out better when they're trained on well-crafted and well-curated synthetic data.
Honestly, everyone seems to assume that machine learning researchers are simpletons who've never used a photocopier before.
fubarx@lemmy.ml 3 months ago
No shit. People have known about the perils of feeding simulator output back in as input for eons. The variance drops off so you end up with zero new insights and a gradual worsening due to entropy.
Zip2@feddit.uk 3 months ago
So it’s basically an AI prion disease?
Llewellyn@lemm.ee 3 months ago
No.
SlopppyEngineer@lemmy.world 3 months ago
Eventually an AI will be developed that can learn with much less data. In the end we don’t need to read the entire internet to get through our education. But, that’s not going to be LLM. No matter how much you tweak LLM models, it won’t get there. It’s like trying to tune a coal fired steam powered car until you can compete in a formula 1 race.
conciselyverbose@sh.itjust.works 3 months ago
Yeah, it’s entirely plausible that LLMs are a small part of the answer as basically the language center of the brain, but the brain is a hell of a lot more complex than that. The language center isn’t your whole brain, and is only loosely connected to actual decision making. It confabulates a lot.
SlopppyEngineer@lemmy.world 3 months ago
OpenAI stumbled on something that worked and ran with it, and people started proclaiming it to be the answer to everything. The same happened with Deep Learning and every AI invention so far. It’s all just another stepping stone on the way.
Even_Adder@lemmy.dbzer0.com 3 months ago
It’s already happening. A quote from Andrej Karpathy :
Turns out that LLMs learn a lot better and faster from educational content as well. This is partly because the average Common Crawl article (internet pages) is not of very high value and distracts the training, packing in too much irrelevant information. The average webpage on the internet is so random and terrible it’s not even clear how prior LLMs learn anything at all.
ConstipatedWatson@lemmy.world 3 months ago
You don’t say, Sherlock
YeetPics@mander.xyz 3 months ago
So do humans if I’m being honest, look at the RNC.
lemmy_get_my_coat@lemmy.world 3 months ago
Can’t wait
FartsWithAnAccent@fedia.io 3 months ago
No shit.
BombOmOm@lemmy.world 3 months ago
Yep. It leads to a positive feedback loop. They just continue to self-reinforce whatever came out before.
And with increasing amounts of the internet being polluted with AI text output…
Ensign_Crab@lemmy.world 3 months ago
… AI inbreeding.
skillissuer@discuss.tchncs.de 3 months ago
hapsburgGPT
Boozilla@lemmy.world 3 months ago
We call it the GRRM model.
LilaOrchidee@feddit.org 3 months ago
AInbreeding
kevincox@lemmy.ml 3 months ago
To be fair this doesn’t sound much different than your average human using the internet.
sp3tr4l@lemmy.zip 3 months ago
2024, Reverse Turing Test Challenge:
Can an LLM AI differentiate between human input and LLM AI input?
MagicShel@programming.dev 3 months ago
That seems so obviously predictable.
Even_Adder@lemmy.dbzer0.com 3 months ago
You have to pretty much intentionally give it enough synthetic data to wreck it. OpenAI and Anthropic train their models on generated data to improve them. As long as there’s supervision during training, this isn’t really a problem.
openai.com/…/prover-verifier-games-improve-legibi…
www.anthropic.com/research/claude-character
Tobberone@lemm.ee 3 months ago
Well… Its built on statistics and statistical inference will return to the mean eventually. If all it ever gets to train on is closer and closer to the mean, there will be nothing left to work with. It will all be the average…