Home / Technology / AI Designs Smarter Chips, Tackling Compute Crisis
AI Designs Smarter Chips, Tackling Compute Crisis
15 Oct
Summary
- AI is being developed to automate and accelerate chip design.
- New AI-designed chips can achieve unconventional, superior layouts.
- Faster chip design could enable custom chips for various applications.
The demand for computing power is projected to increase by 10,000x in the next five years. Recursive Intelligence founders Anna Goldie and Azalia Mirhoseini are developing AI to design chips more efficiently, aiming to mitigate this impending compute crisis. Their AI system learns from each design iteration, continuously improving subsequent ones.
This AI-driven approach challenges traditional chip design by discovering novel layouts. For instance, their AlphaChip system previously generated a successful circular component arrangement. Instead of partitioning chips into smaller blocks, Recursive Intelligence optimizes entire chips, enabling intricate designs.
Goldie posits that AI handling execution will free humans to focus on creative ideas and applications. This AI-assisted design process could drastically reduce chip design timelines from months or years to as little as a week. Such speed would make custom chip development financially viable for diverse applications like AR, VR, and biological computing, all while improving energy efficiency.
