
Exhaustion of High-Quality Data Could Slow Down AI Progress in Coming Decades
The Data Exchange with Ben Lorica
00:00
Do You Track Friends in Memory?
There's also some more slow in many ways is focus on flops, but there's also memory, right? So memory bandwidth interconnects because at some point these models get too so big that you really need to do distributed computing. We are working on that, but it's not done yet. But yeah, I would say it's fairly likely that we will see more compute architectures that use processing in memory and things like that.
Transcript
Play full episode