Skimming through the article, it seems like the extent of this is to require IAAS (Infrastructure) providers to verify the identity of those who are using their services to train AI. It's an attempt to stymie sanctioned or malicious actors, from training AI and especially from hopping between services or using aliases to continue training on their model. It seems a bit benign and I don't understand the parallels othe…
From the executive order (Executive Order 14110) it seems to affect only massive compute infrastructure: > (i) any model that was trained using a quantity of computing power greater than 10^26 integer or floating-point operations, or using primarily biological sequence data and using a quantity of computing power greater than 10^23 integer or floating-point operations; and > (ii) any computing cluster that has a set…
Something like 40 of them, or 100-300 if you're looking at FP16. So well over 2^14.
And that's per second, give it your idle cycles for four months and that's 10^7 seconds.
It gets pretty close to 10^23.