Overview

• 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.