AI Social + optional discussion about data augmentation for LLM training

Come and chat with other folks interested in generative AI.

Optional discussion topic: If we had a largish fact checked subset of the web, would we be able to train better models faster?

Optional research paper reading: If you are interested in the technical side of AI, we will have an optional discussion of the recent paper: "Rephrasing the Web: A Recipe for Compute and Data-Efficient Language Modeling". Translating web page text to neutral point of view, wikipedia style prose dramatically improves training speeds.

https://arxiv.org/pdf/2401.16380.pdf