PyTorch Conference China: open source is key to coordinating hardware, models and inference
PyTorch · x · 2026-10-01
At PyTorch Conference China 2026 in Shanghai, PyTorch Foundation Executive Director Mark Collier delivered the keynote "Building Frontier Intelligence in the Open."
His core argument: as AI hardware, model architectures, and inference engines evolve rapidly, the real challenge is coordination across the three — and the flow of information is the potential bottleneck. "The faster the information can flow, the faster the progress can be made, and that's exactly where open source really shines."
As a concrete example, he cited Shopify running agents in production, where failures feed back into training with PyTorch and serving with vLLM as models improve. The full keynote is online; PyTorch Conference North America runs October 20–21 in San Jose.
Related event: PyTorch Conference China: Open Source as Key Coordinator for AI Progress(3 posts)→
More from Companies & People
- Perplexity design lead previews ongoing brand refresh at Budapest's POV design conference — pklada · 2026-10-01
- OpenAI CFO Sarah Friar takes CNBC questions comparing Dots to Muse — rohanpaul_ai · 2026-10-01
- Professor: AI cut 80 hours of Canvas drudgery per semester, freeing time for teaching — _akpiper · 2026-10-01
- Lenny shares PM talk: the craft of product management is decision-making above all — lennysan · 2026-10-01
- AI researcher Timnit Gebru named 2026 Right Livelihood Laureate for challenging big tech power — MilagrosMiceli · 2026-10-01
- VC Khosla slams 'unethical' AI rival; Kaplan notes he invests in both Cognition and Factory — russelljkaplan · 2026-10-01