I mean, even AI isnt as bad as people make it out to be, people have GROSSLY overexaggerated the consumption of it to fit their narratives
If you crunch the numbers, a few hours of using even the most hungry of LLMs uses less water than washing your hands once >_>;
People don’t wanna hear it though because it doesnt fit their “AI bad” narrative
Millennials need to take a very long look in the mirror and realize this is their “phones bad” boomer moment and take the initiative to grow up. Its cringe and most of their gripes arent backed by reality.
Do LLM datacenters eat up power? Are they bad for the environment? Do they use a lot of water?
Yes… but they arent even in the top 10 of worst offenders that have been around for way longer and per-capita fuck shit up way worse
My fellow Millennials are only hating on AI cuz they are scared of it, or deeply deeply misinformed.
Id honestly be really curious to see a pool on how much water people think 1 convo with an LLM actually takes, Id bet the answers are so incredibly far off.
(The reality is its only a couple mL of water or so, depending on the model, if you wanna know)
I am not convinced LLMs provide a damn thing of value.
I work as a software developer and the sheer increase to my productivity makes it extremely valuable for me on numerous fronts.
Id estimate Im about 4x as productive, give or take.
And most of that productivity isnt coding, most of it is what LLMs are good at:
Locating documentation (using it as a fuzzy find searcher by intent)
Generating documentation and improving documentation
Summarizing existing code
Quickly locating bugs
Its really good at fuzzy searching huge datasets, finding needles in haystacks. When Im trying to find something specific in a codebase with thousands of files quickly.
While I agree most people are using it wrong or poorly, if you do know how to use an LLM, it becomes a huge productivity boost.
I mean, even AI isnt as bad as people make it out to be, people have GROSSLY overexaggerated the consumption of it to fit their narratives
If you crunch the numbers, a few hours of using even the most hungry of LLMs uses less water than washing your hands once >_>;
People don’t wanna hear it though because it doesnt fit their “AI bad” narrative
Millennials need to take a very long look in the mirror and realize this is their “phones bad” boomer moment and take the initiative to grow up. Its cringe and most of their gripes arent backed by reality.
Do LLM datacenters eat up power? Are they bad for the environment? Do they use a lot of water?
Yes… but they arent even in the top 10 of worst offenders that have been around for way longer and per-capita fuck shit up way worse
My fellow Millennials are only hating on AI cuz they are scared of it, or deeply deeply misinformed.
Id honestly be really curious to see a pool on how much water people think 1 convo with an LLM actually takes, Id bet the answers are so incredibly far off.
(The reality is its only a couple mL of water or so, depending on the model, if you wanna know)
How about the water usage of energy for data centers? Or for that each query there’s tons of pre and post training usage?
How about the fact that it’s just a grift that doesn’t do much and even environment aside it funnels wealth to the richest people?
Hank green has a nice little video of it
https://m.youtube.com/watch?v=H_c6MWk7PQc
Yes, I included that.
The usage is incredibly low “per capita”
If you think AI is special, its not. This same problem is present when you watch Netflix or order from Amazon.
People have a very warped idea of how much power/water AI uses.
You should go look up how huge the datacenters that power everything else are too.
But again, like Netflix provides something now (albeit not my favorite thing). I am not convinced LLMs provide a damn thing of value.
I think AI is special for being a waste and very negative socially without providing any positive back.
I work as a software developer and the sheer increase to my productivity makes it extremely valuable for me on numerous fronts.
Id estimate Im about 4x as productive, give or take.
And most of that productivity isnt coding, most of it is what LLMs are good at:
Its really good at fuzzy searching huge datasets, finding needles in haystacks. When Im trying to find something specific in a codebase with thousands of files quickly.
While I agree most people are using it wrong or poorly, if you do know how to use an LLM, it becomes a huge productivity boost.