Maybe they new that, and entire generations of computer games mobilised, markets created to fund the iterations they knew could take 40 years?

Not as crazy as using 30 Playstation 3s to power a super telescope cause it was cheaper

It kind of went the other way around. Big vector processing has been around for year (i.e., Cray) s. It was used mostly for simulation and cost a fortune. GPUs took the idea and comoditised it making it affordable for gamers. 1/2

Its been a wild ride I remember back in the day building 3DFX dual SLI gaming rigs for rendering games, now the same tech has evolved to render digital twins environments, neural nets, and more.

Parallel processing was the key. GPUs are designed for parallel processing while CPUs for serial. Matrix processing is parallel.

Maybe there’s a way better solution that no one ever explored because all the programmers were big gamers and looked for a way to use their gaming PCs for it

Whenever there is an excess of something, a tech rises to utilize it. Storage, processing power, bandwidth, etc.

Greg do you think there is any chance of bringing Open AI five back in a way that could help new players learn Dota 2? Current bot logic is awful and real match making is rough for everyone in the game

Now that everyone has experienced AI and its massive utility to society, do you think manufacturers might be compelled to create a new processing architecture that might reduce the cost and electricity consumption and simultaneously be more efficient in contributing towards AGI?