Open Innovation: Unlocking AI's Potential
summary
Argues that open innovation drives faster progress in models and systems, citing an actual Mixture-of-Experts model — dbrx mixture-of-experts trained with PyTorch and community collaboration — released as open source and reused by others. Lists real tooling that matters: LangChain, LlamaIndex, AutoGPT, Milvus, FlashAttention, DeepSpeed, SG Lang, TGI. Gives concrete deployment math: Llama 3.1 405B needs 8 H100s (~405 GB), ~16 petaflops, ~$50/hr ≈ $400k/yr for one 2k-token/sec endpoint.
AI-written summary · about
chapters (16)
- 0:00Introduction
- 0:36AI Summer
- 0:56Growth
- 2:04History
- 4:16Personal Experience
- 5:21The Future
- 6:30Key Requirements
- 7:45Hot Take
- 8:22AI Systems
- 9:32Open Source
- 10:26Highly Capable Models
- 11:31My Personal Experience
- 12:51Open Foundation Models
- 15:19Open Innovation Projects
- 15:53SGLang
- 16:31Conclusion
related talks
54mThe Future of AI: Building the Most Impactful Technology Together
16mKeynote: Open Source for AI and AI for Open Source
27mKeynote: Open Source in the Age of AI
01h 08mPanel Discussion: The Impact of AI Regulation on Open Innovation in AI
12mCo-Evolution: How the Open Source Intelligence Stack Compounds
22mThe Imperative for Open Source AI - The Path Forward to Take Down the AI Oligarchs
22mBeyond boundaries: How global collaboration defines AI in 2026
58mAI in Reality Fireside Chat: Enterprise AI & Open‑Source Innovation