Ricursive Intelligence develops AI system to automate chip design
publishers 84articles 225first reported 23 Sep, 14:00 UTCdeveloping since 30 Sep · 32 editions
Ricursive Intelligence is developing artificial intelligence that can design entire chips, learn from each iteration and improve later designs, its founders Anna Goldie and Azalia Mirhoseini said at TechCrunch Disrupt.
The startup aims to automate and accelerate a process that can currently take months or years. Goldie said rising demand for computing cannot be met simply by building vastly more data centers, arguing that more efficient chips could make better use of bandwidth, semiconductor fabrication capacity and energy.
The founders said AI can produce unconventional layouts that outperform traditional designs. They cited AlphaChip, a system they previously developed, which created a circular arrangement of components that initially surprised engineers but proved successful. Ricursive is working to optimize whole chips rather than dividing them into dozens of smaller blocks, an approach the founders said enables more intricate layouts.
Goldie said shortening design cycles to roughly a week could make custom chips financially practical for a wider range of uses, including space technology, augmented and virtual reality, and biological computing. She argued that delegating chip-design execution to AI would allow people to concentrate on applications and new ideas.
HOW THIS STORY WAS MADE
Written from 8 articles, 7 with the publisher's own text; 8 independent newsrooms once syndicated copies count as one; this version written 354 h after the record first saw the story; the editor kept it as written.
- Publishers
- 84
- Source articles
- 225
- Given to the writer
- 8, 7 with the article's own text
- Independent newsrooms
- 8
- This version written
- 354 h after the record first saw the story
- Second model (editor)
- kept as written
- Publication gate
- passed
- Human review
- none
WHAT THE EDITOR DID
The second model checked the writer's draft against the same sources and published it as written.
Written by a language model from the sources above, then checked by a second model that may only cut, attribute or correct. How it works →