Comment by kennywinker
2 days ago
https://sustainability.atmeta.com/blog/2024/10/14/our-approa...
Anthropic: no sustainability goals or claims
OpenAI: no sustainability goals or claims
SpaceX AI: no sustainability goals or claims
Some other non-lab inference providers:
https://regolo.ai/ Eu based, 100% renewables powered
https://greenpt.com/us/sustainability Eu based, allows you to track carbon cost of inference.
Acting like there is no difference between bad things is how you get worse things.
The fact that you believe Meta, of all companies who's gone to great lengths to lie about their core business model, speaks volumes.
This also highlights my point: the norm is that the bulk of the inference and training providers have no goals.
You've moved the goalpost and declared victory my guy. I literally gave you an option, that isn't meta, and is run on renewable energy. The fact that it's hard to get clear information doesn't negate the fact that there are differences, and we should seek to find the best option rather than giving up and accepting whatever bad option is nearest or shiniest.
Let me clarify my perspective here. You put the frontier providers at the top. That's probably representative of >80% of the total, global, inference. The other ones you mention? They're likely showcasing "green AI" as a differentiator, but comparatively, they likely represent a percent of global capacity.
And you don't get to put a feel good landing page from Meta and "declare victory" (I didn't) just the same. It means nothing because there isn't any accountability. Nice try.