@nishu

all the ai companies are looking to now accelerate along the lines of hardware now as that's the bottleneck for now before they jump back into improving the algorithm they want to get to bare bones elon, sama, google → custom gpu's the problem they are really trying to solve is not ai learning but memory how do you compress large pieces of information into smaller pieces of information it's a compression problem learning is seeing large data chunks and compressing into smaller repeatable patterns ok so hardware has limits or atoms have limits, you could go subatomic etc., but is there any element that can pack lots of information in small place for less energy? how do we/humans do it? cell? cellular compression as a means to get over atomic compression then i found this yt of this startup using human neurons for machine learning tldr; cellular compression is the next one in line [i feel] https://www.youtube.com/watch?v=wBtr8iv7onA https://x.com/elonmusk/status/2059884150187053488?lang=en

1 reply

0 recast

1 reaction