this post was submitted on 05 May 2026
53 points (96.5% liked)

technology

24367 readers
320 users here now

On the road to fully automated luxury gay space communism.

Spreading Linux propaganda since 2020

Rules:

founded 5 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
[–] KnilAdlez@hexbear.net 20 points 1 week ago (3 children)

It's strange to me to call out the climate costs, especially since this would have far less climate impact than having to call to google's servers to use the full sized models. LLMs aren't magically worse for the environment, its the hardware they run on for the full sized models that is incredibly power hungry. A 4 GB LLM would probably use less power than a modern video game on a person's computer, and run for less time.


[–] AstroStelar@hexbear.net 11 points 1 week ago* (last edited 1 week ago)

I was thinking the same thing. The article claims 30,000 to 60,000 tons of CO2e emissions from sending 4GB of data to hundreds of millions of phones. For reference, estimates for the US AI industry's emissions are between 30 and 80 million tons per year, global total emissions are 80 billion. How often this gets repeated for new versions is unclear.

As for inference, Chrome won't even use it for its biggest use case: everything done via the search bar and "AI mode" is still sent to high-param models in Google's servers, likely because user data is Alphabet's cash cow. The local model is only used in very niche cases:

the features that do use the local model (Help-Me-Write in , tab-group AI suggestions, smart paste, page summary) are buried in textarea-context menus and tab-group right-click menus that the average user will discover, on average, never.

[–] chgxvjh@hexbear.net 9 points 1 week ago (1 children)

LLMs aren't magically worse for the environment, its the hardware they run on for the full sized models that is incredibly power hungry.

Not magically, just technically. Within a couple years we just made computers use significantly more energy for no good reason. This shit being into everything is incredibly unpopular.

[–] KnilAdlez@hexbear.net 6 points 1 week ago (1 children)

I'm not sure if I understand you, it sounds like you think running a small llm locally on your computer will suddenly make it use like 10x more power. That's not how it works. It's the servers used to run the full sized models that use that much power, as each one has tens of thousands of processors running at once. And local llms do have usage, especially for accessability. I use a local llm for my home assistant instance so I can use voice commands, which is very helpful as a disabled peraon.


[–] chgxvjh@hexbear.net 3 points 1 week ago (1 children)

You seem to have no idea how good modern computers are at idling

[–] KnilAdlez@hexbear.net 7 points 1 week ago (1 children)

What does that have to do with anything?

[–] NewOldGuard@lemmy.ml 1 points 1 week ago (1 children)

I think their point is that regular web browsing will use less power than web browsing with local LLM calls. Your PC running an LLM is likely gonna hit its TDP limits, while browsing will be a fraction of that. Yes it’s less power than used by a trillion parameter model but I think their point is it’s vastly more than your non-LLM standard browsing would be

[–] KnilAdlez@hexbear.net 3 points 1 week ago

Your PC running an LLM is likely gonna hit its TDP limits

Debatable for a 4GB model, depending on the hardware. It's also (most likely) not constantly running, so while yes, it will use more power than not having it, whether or not it is a significant change in the long run depends on many factors.


[–] GaveUp@hexbear.net 4 points 1 week ago

It's an easy angle to win over the unaware general public by hitting points like climate and water usage

Dishonest, and a little silly to people who understand the math but it is an angle