Keynote: Open Source for AI and AI for Open Source
summary
Argues that open models and tooling make enterprise AI controllable while AI reshapes open source by generating code and tests; cites a fivefold performance gain from multi-team open collaboration, production projects like Granite and LLMD (transferred to CNCF) that deliver 2–3x throughput improvements and lower latency, and a custom ASIC, IBM Spire, integrated via PyTorch/Open RAG. Reports a concrete security win: Glasswing plus local models finds a confirmed vuln missed by Mythos, with exploit and patch produced.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 1:04From models to systems
- 3:07Open source's role in AI
- 5:14IBM and open AI models
- 8:00Open model diversity
- 9:01Deploying models in production
- 10:24LLMD for distributed inference
- 12:08Accelerators and ecosystem
- 13:34AI impact on software and security
- 14:56Security tools and open source future
related talks
13mKeynote: Introductory Remarks
27mKeynote: Open Source in the Age of AI
22mThe Imperative for Open Source AI - The Path Forward to Take Down the AI Oligarchs
16mKeynote: Open Source AI Everywhere
13mThe Good, The Bad And The Ugly Of Open Source AI
13mKeynote: State of Open GenAI- 13mKeynote: State of Open GenAI
42mHow Open Standards Can Heal the Open Wound of AI