Keynote: Tracking the Pulse of Generative AI
summary
Claims access to 30+ models with most real use concentrated on OpenAI, Anthropic, Gemini and Llama; survey data shows healthcare budgets: over a third plan to double GenAI spend and nearly half increase by 50%. Recommends focusing on post-training / fine-tuning (quantization, distillation, RLHF) rather than training from scratch. Describes practical patterns—Routers and ensembles, cascades and model merging (mergekit), embedding-with-context (Voyage AI)—and production tools Ray, VLM, Sycamore and graph RAG examples including a Knowledge Graph of 240 billion data points.
AI-written summary · about
chapters (13)
- 0:00Introduction
- 0:38Foundation Models
- 5:10Challenges
- 7:25Posttraining
- 17:52Examples
- 20:14Structure
- 22:19Agents
- 25:20Analysts
- 28:20Infrastructure
- 30:12Ray
- 31:14Multimodality
- 35:36Building AI Applications
- 36:44AI Risks
related talks
11mKeynote: Building an Interoperable Ecosystem for Generative AI
17mGenerative AI: Keeping it real
01h 20mGenerative AI for Businesses
26mGenerative AI: Get a Grip … and a Look Under the Hood
43mKeynote: Applied NLP in the age of Generative AI
13mKeynote: State of Open GenAI
47mThe Rise of AI: A Reality Check on Generative AI
19mLet's Talk GenAI