@danjones000 @Matt_Noyes read the rest of the thread.
Uncategorized
9
Posts
3
Posters
3
Views
-
@danjones000 @Matt_Noyes read the rest of the thread. Adding in training gets us to about 0.0065kg CO2eq, although it's not clear if they count that in the original estimate.
-
@danjones000 @Matt_Noyes read the rest of the thread. Adding in training gets us to about 0.0065kg CO2eq, although it's not clear if they count that in the original estimate.
This post is deleted! -
@danjones000 @Matt_Noyes read the rest of the thread. Adding in training gets us to about 0.0065kg CO2eq, although it's not clear if they count that in the original estimate.
@evan@cosocial.ca The article that @Matt_Noyes@social.coop posted in the thread pretty clearly laid out the massive amount of electricity that AI data centers are using. That article isn't the only one saying the same thing. AI is using massive amounts of power.
Traditional server racks consume 5-15 kW, while AI-optimized racks with high-performance GPUs require 40-60+ kW. Some cutting-edge AI training facilities are pushing individual racks to 100+ kW, fundamentally changing data center design and cooling requirements. (ref)
It doesn't really matter if the power is coming from requests to the API, running the models, training the models, or making ritual sacrifices to Baphomet in hopes of making the models sentient.
If someone is using AI, they are indirectly contributing to that power usage. If you can acknowledge that and make peace with it, fine. But, saying that the energy cost is minimal in light of this is ignoring reality.
-
@evan@cosocial.ca The article that @Matt_Noyes@social.coop posted in the thread pretty clearly laid out the massive amount of electricity that AI data centers are using. That article isn't the only one saying the same thing. AI is using massive amounts of power.
Traditional server racks consume 5-15 kW, while AI-optimized racks with high-performance GPUs require 40-60+ kW. Some cutting-edge AI training facilities are pushing individual racks to 100+ kW, fundamentally changing data center design and cooling requirements. (ref)
It doesn't really matter if the power is coming from requests to the API, running the models, training the models, or making ritual sacrifices to Baphomet in hopes of making the models sentient.
If someone is using AI, they are indirectly contributing to that power usage. If you can acknowledge that and make peace with it, fine. But, saying that the energy cost is minimal in light of this is ignoring reality.
@danjones000 @Matt_Noyes electricity is only about 1/3 of global emissions. All data centers, including AI datacenters, are only 1% of electricity usage. That makes 0.3% of total emissions.
Much more emissions are due to cars, meat, cement production and rice cultivation.
I recognize that AI has a lot of electricity use; it is nothing compared to the other things you do with your time.
-
@danjones000 @Matt_Noyes electricity is only about 1/3 of global emissions. All data centers, including AI datacenters, are only 1% of electricity usage. That makes 0.3% of total emissions.
Much more emissions are due to cars, meat, cement production and rice cultivation.
I recognize that AI has a lot of electricity use; it is nothing compared to the other things you do with your time.
@danjones000 @Matt_Noyes It does not hurt to try to reduce those emissions; reducing any emissions is good.
But claiming that AI is solely responsible for climate change is intellectually dishonest.
There are plenty of other problems with AI; "burning up the planet" is not a convincing one.
-
@danjones000 @Matt_Noyes It does not hurt to try to reduce those emissions; reducing any emissions is good.
But claiming that AI is solely responsible for climate change is intellectually dishonest.
There are plenty of other problems with AI; "burning up the planet" is not a convincing one.
@evan @danjones000 This does sound like denialism, though, doesn't it? Cow farts, and all.
-
@danjones000 @Matt_Noyes It does not hurt to try to reduce those emissions; reducing any emissions is good.
But claiming that AI is solely responsible for climate change is intellectually dishonest.
There are plenty of other problems with AI; "burning up the planet" is not a convincing one.
@evan @danjones000 Respectfully, the "shaming people" line is not serious, right? "solely responsible for climate change" - had anyone ever made that claim?
-
@evan @danjones000 Respectfully, the "shaming people" line is not serious, right? "solely responsible for climate change" - had anyone ever made that claim?
@Matt_Noyes @danjones000 I don't think climate change is a fair last resort when questioning someone's use of LLMs. There are plenty of other reasons not to use them.
-
@evan @danjones000 This does sound like denialism, though, doesn't it? Cow farts, and all.
Denying anthropogenic climate change? No. I worked for four years in a climate change non-profit. In this thread, I gave a few examples of human activities that cause greenhouse gas emissions.
My sole point is that it's sad when people don't understand the relative importance of different activities for causing climate change.
Gli ultimi otto messaggi ricevuti dalla Federazione
-
Fuuuuuuuuuuuuuuuuuuuuuuuu
-
Trump manda nave sanitaria in Groenlandia
-
@jennyrozema wat super naar :(
-
@evan who are playing?
-
@eniko Yeah, that one is *really* expensive even on simple geometry. Unless you have some shading magic that doesn't work in Eevee, you probably should give that one a shot - it's a lot closer to "real time".
-
@ratsnakegames cycles
-
@Bishop @HurricaneWeasel @gmarcosanti @borgqueen
Ha vinto il blu, anche perché mi faceva fatica fare il cambio di nuovo 🤣
-
@snow avevo letto del progetto LoRa, infatti mi sarebbe piaciuto realizzare una mia stazione, però non lavorando evito di spendere soldi per una cosa che non mi serve
