How Ricursive Intelligence’s Founders are Using AI to Shape The Future of Chip Design
- Source
- youtube.com
- Author
- Sequoia Capital
- Date
Context
Anna Goldie and Azalia Mirhoseini, founders of Ricursive Intelligence, say chip design is slow enough to hold back AI hardware, and that AI could shorten it. In a Sequoia Capital podcast they describe a mismatch: new AI methods appear much faster than new chips can be designed, which limits how well chips and models can be tuned to each other.
Their evidence is earlier work on AlphaChip at Google, where they trained a reinforcement learning (a that improves by trying options and learning from results) to place components on a chip. They say it learned from experience, produced unusual curved layouts that can shorten wiring, and that a chip using it was built and worked. They add that Google's TPU team was at first sceptical and wanted routed wire length, congestion and timing, not the academic measure first reported.
Ricursive aims to extend this from placing components to the whole design flow, relying on synthetic data. That plan is a stated goal, and the result claims come from the founders.
- Agent — A system in which a model repeatedly chooses operations, a harness carries them out, and the results return to the model to inform its next choice.
- Program — A set of instructions stored so that a computer can run it.
- Model — A computational system that uses learned patterns to produce an output from information supplied to it.
videoBringing Continual Learning to Every Company: Trajectory’s Ronak Malde
videoCrusoe CEO Chase Lochmiller: Turning Stranded Energy Into Intelligence
videoDavid Silver and Lasse Espeholt of Ineffable Intelligence: A Superlearner for the Age of Experience
videoCracking the Code on Offensive Security With AI ft XBOW CEO and GitHub Copilot Creator Oege de Moor
Checking sign-in…
Loading comments…


