https://twitter.com/WSJTech/status/1755774564896756101

This feels very counterproductive?

Like, from an "AI safety" perspective, scaling current techniques to even bigger compute -> earlier superintelligence -> more risk.

And from an "AI openness" perspective, it increases the divide between megacorps and plebs.

There are luckily some rate limiters to AI infra growth: Land, power, water, roads, permitting, transformers, switchgear, etc. The question is if the “AI openness” movement can organize and deploy infrastructure capital faster than the Investment Grade balance sheets can.