Ayar Labs Bets on Optical I/O for Next-Gen Computing
As the dreams of artificial intelligence and machine-learning system developers have grown ever larger, the suppliers of hardware to enable those systems have started to bump up against some fundamental constraints. For example, Nvidia—a leader in graphics processing units (GPUs) that do the heavy lifting for a wide range of computationally intense systems such as those in AI—believes that, by 2023, advanced AI models could entail 100 trillion or more connections, rivaling the number of synapses in the human brain.
