Lightmatter and Lightelligence Showcase Photonic Chips for AI Processing
The new chips use light to perform computations with greater speed and energy efficiency, marking a step toward commercial viability despite challenges ahead.
- Lightmatter and Lightelligence have unveiled photonic chips capable of performing AI tasks like matrix multiplication and running deep learning models with high precision.
- These chips leverage light instead of electricity, offering significant improvements in speed and energy efficiency over traditional electronic processors.
- The technology addresses limitations of Moore's Law and can be manufactured using existing semiconductor facilities, enabling scalability.
- Demonstrations include solving optimization problems, playing Atari games, and executing AI models like BERT with performance comparable to conventional chips.
- While prototypes are nearing practical applications, experts estimate that mainstream adoption of photonic computing may take up to a decade.
Tech Xplore 
New Scientist 
Australian Broadcasting Corporation 
Science News 
Interesting Engineering 
Study Finds 