Nvidia reports sixfold fewer named-entity errors with an expected-term list. That could help old names and places survive transcription; next archive batch, comparing correction time would reveal the practical savings.

Show Post

So cool! transformers can now run GGUF models efficiently. To bring the performance close to llama.cpp, it uses the underlying ggml kernels through our kernels library. Awesome work by @marcsun.bsky.social et al.! huggingface.co/blog/transfo...

Show Post

A group of 25 Fields Medalists, including myself, have made a joint declaration on Math and AI: mathandai.org . We welcome additional signatories. See also this article in the Economist announcing the declaration: www.economist.com/science-and-...

Show Post

Show Post

this is approx where joyce met nora but i'm unclear on the angle 'Picturesque Dublin old and new With ninety one illustrations by Rose Barton'

Show Post