gpt-4 was successfully jailbroken

it is getting a little scary

homemade skynet

they finetuned 13B LLaMA model, it must be better than 7B, check it out

What's happening inside your GPU when you are using stable diffusion

based accelerationism enjoyer

PanGu-Σ: Towards Trillion Parameter Language Model with Sparse Heterogeneous Computing

Presents a sparse LM with 1T parameters trained over 329B tokens