[xp84]

What’s built with all that VC money is already built though; I don’t foresee a future a few years out where we don’t have access to an open-source model roughly as good as the current flagship models for the cost of the compute itself.

[ygouzerh]

It's like the rail industry analogy: we got a big bubble, but the rails are still there. Now with llm, we can just distill expensive one to create cheap open-source ones indeed

[LurkandComment]

Variable costs - electricity etc. Current model is very resource intensive. You know when they build all those Olympic Venues and then once the Olympics is done the ongoing cost is too expensive and then they become derelict buildings.... like that...

[cm2012]

Training is resource intensive. Serving gpt4o is not