The Hardware vs Software
summary
Debate centers on whether compute becomes a utility or stays decisive: panelists cite compute doubles every 3.4 months and MLPerf as the concrete meter of progress, plus a reinforcement‑learning floorplanning paper and Oxide’s rack‑scale prototype as real artifacts. One side points to M1’s performance‑per‑watt and Cerebras wafer‑scale chips to argue hardware–software co‑design yields order‑of‑magnitude gains; the other side points to APIs, transfer learning and open frameworks to argue abstraction already hides hardware and scales AI broadly.
AI-written summary · about
chapters (17)
- 0:00Introduction
- 3:19ML Commons
- 4:10Autodesk
- 4:53Anna Goldberg
- 5:34Brian Cantrell
- 22:10Moores Law
- 29:09Crypto
- 32:18Hardware vs Software
- 36:15The Rise of Rust
- 39:08Davids Response
- 41:45Conclusion
- 43:58Power of abstraction
- 46:09Bring it home
- 47:57Special Chat
- 52:05What matters for ML
- 54:06Imagenet
- 56:07General Purpose FPGA
related talks
58mThe Next Decade of Software Development
24mIndustry Panel
01h 09mThe Future of Software Engineering
15mKeynote: From Platforms to AI Factories. Has Kubernetes Solved It?
05mOpening Remarks
25mMaking AI Useful
32mBuilding Once, Running Everywhere: The Future of Portable AI Applications
24mSoftware 3.0 and the AI Engineering Landscape