Seeing Like a Language Model - Understanding Embeddings and Meaning in AI

Explore the fascinating world of language model perception in this 34-minute conference talk by Linus Lee, Research Engineer at Notion. Delve into recent breakthroughs in interpretability research to understand how embeddings represent meaning and how language models process text. Discover encouraging updates on Lee's ongoing exploration of these topics and learn how these insights can be applied to improve retrieval-augmented LLM systems and create more intuitive interfaces for reading and writing. Gain valuable knowledge from Lee's experience in prototyping AI-augmented tools for thinking and collaborative work. This talk, presented at the AI in Production Conference by MLOps.community, offers a unique perspective on the inner workings of language models and their potential to enhance our creative and collaborative processes.