Post #2937825
2026-05-14 04:44 UTC
@issackelly@mastodon.social Yeah... among many aspects of it... it's hard to separate LLMs as a _general technology_ and what they could or will do from the way it's actually implemented currently...as well and the resources involved. A whole lot of talk about "efficiency" and we are talking about [checks notes] a $600B spend schedule on AI buildout from just one of those companies.
Has anyone done the actual math? This is like saying I'm an effient beer drinker when I'm running up someone elses tab.
Replies (1)
-
@issackelly@mastodon.social 2026-05-14 04:50
@cda@macaw.social yeah! I am in a little discord where there are some of these experiments going on with the open models and agents and accessible or rentable hardware. Trying to get a little bit of “dollars and watts per …. Some kind of output” without the VC subsidies. There is some promising output, but it’s a load of hardware, and the cost of acquiring hardware is getting complicated.