Keynote: Teaching LLMs New Knowledge
summary
Claims an all-in-one stack that adapts any model with a few lines of code and removes costly data collection/transformation, while adding hundreds of compute nodes for users. Demonstrates concrete tools: programmatic JSON outputs (example: force an INT and units field to avoid brittle string parsing) and workflows that ingest millions of user data points to teach domain facts, correct wrong knowledge, and change behavior via fine-tuning, pretraining, and RAG combined for durable model updates.
AI-written summary · about
related talks
05mKeynote: Teaching LLMs New Knowledge
12mKeynote: How LLMs Have Changed the Way Organizations Do AI
08mKeynote: Plug in and Scale: Serving LLM Models on Kubernetes Ma...
16mKeynote: An Open-Source Stack for AI Compute: Kubernetes + Ray + PyTorch + vLLM
33mOpening Keynote: LLMs, between fantasy and reality - the poin...
44mKeynote Speaker - Simon Willison
33mKeynote: Fine-tuning our way towards openness in AI
16mKeynote: DBRX: Training Modern LLMs on Databricks