A small number of samples can poison #LLMs of any size
"Specifically, we demonstrate that by injecting just 250 malicious documents into pretraining data, adversaries can successfully backdoor LLMs ranging from 600M to 13B parameters."
cc @asrg
destroyai
"“It’s extremely hurtful, frankly,” said Nvidia chief executive Jensen Huang...OpenAI CEO Sam Altman has concurred, lamenting pushback... “Looking at what’s possible, it does feel sort of surprisingly slow"
We're making the AI broligarchs sad, friends, but we can do better.
MOAR SADDNESS FOR SAM AND JENSEN
#DestroyAI :hammer:
https://futurism.com/future-society/tech-ceo-ai-hate