4 ms·
>Could a "GPUs too cheap to meter" phase—say, about a decade, up to 2040—remarkably speed up cycle times of traditional deterministic modeling / simulation type
by fuzzfactor 9mo ago
>Could a "GPUs too cheap to meter" phase—say, about a decade, up to 2040—remarkably speed up cycle times of traditional deterministic modeling / simulation type workloads.
Seems to me for it to really go wild the bottleneck to overcome would need to include GPU experts too cheap to meter also.
- adityaathalye 9mo agoThe people assembling GPU boxen would need work. Big Datacenter will likely turn off the power and evict those human cost centers, as a first step toward asset liquidation. Which is why I hope that occurrence feeds forward into a glorious SME business boom. e.g. Nokia's mobile telephony self-own by the end of the aughts was both sad and great for Finland; hurting national pride at one end and fuelling their high tech startup scene, exactly due to losing experts to attrition and entrepreneurship.