Light-Based Chips Could Help Slake AI’s Ever-Growing Thirst for Energy. Optical neural networks, which use photons instead of electrons, have advantages over traditional systems.
Light-Based Chips Could Help Slake AI’s Ever-Growing Thirst for Energy. Optical neural networks, which use photons instead of electrons, have advantages over traditional systems.
Optical neural networks, which use photons instead of electrons, have advantages over traditional systems. They also face major obstacles.
Also making these companies pay the same electricity rates as the average citizen could help too...no more company rebates to syphon everything for 1/100th the cost
TL;DR: Still no "optical transistor" yet, but they figured out a way to do matrix multiplication:
The researchers encoded the various quantities they wanted to multiply into beams of light, then sent the beams through a series of components that altered the beam’s phase—the way its light waves oscillated—with each phase alteration representing a multiplication step. By repeatedly splitting the beams, changing their phase, and recombining them, they could make the light effectively carry out matrix multiplication. At the end of the chip, the researchers placed photo detectors that measured the light beams and revealed the result.
Was going to say that phototransistors have existed for 70 years https://en.m.wikipedia.org/wiki/Photodiode
But... Then realized that you said optical (light as medium, not light controlling electricity).
Does that that means it scales nicely with bigger matrices or do you still need to do more operations ?
I'm not sure how this would scale with larger matrices, and the article is a bit thin on the details.