Natural Optical Computing

Multisynapse optical network outperforms digital AI models

This one experiment proves we will soon be able to replace all the graphics cards in Meta's new computer, covering 2,250 acres, with something the size of a breadbox, that works at the speed of light, using maybe 3 watts.

This is the very last piece of the puzzle required to design a God-Like AI, roughly a million times faster than anything today. Rather than relying solely on quantum computing, it can produce results similar to what graphics now do, not because graphics cards are efficient, but because they're the cheapest way to go.

Supercomputers sell their old graphics cards, and this technology will float to the top first. I've given up trying to predict exactly when these things will happen, but graphics cards are rapidly becoming history, with AMD's new ALL-IN-Wonder chips blowing everyone's minds. The newest chips can have almost 200 processors and 350gb of ram, as well as, four gpus. Theoretically, all that can be replaced with as few as 1,500 optical transistors, that work a million times faster, using legacy bargain bin chips around the 28nm and up, that can also efficiently talk to electronic circuitry.