91–100 of 125 posts
I think AI is interesting, but I wonder what it will do to progress when used very extensively. Trained on all the things/ideas of the past and creating a larger barrier of entry for new ideas. Thinking about new computer languages, who would use one that AI couldn’t help you code in?
> Firstly, there’s the environmental impact. Surely, like everything else in tech, this too shall pass. I expect power requirements to fall away since there is no doubt strong incentives to do so. > My naive optimism led me to believe that technology would help us fight climate change. Yeah, proceeding full steam ahead with planet destruction and praying tech will save us is kind of naive. You're not alone though. >…
> The web has been contaminated ever since SEO. Maybe AI will kill the web. So it goes. This statement kinda blows my mind. You contemplate the destruction of the web, the single most world-changing invention of the last century, possibly the last millennium, a force that has democratized access to information, accelerated scientific progress and connected people around the world, the technology that one way or anoth…
It may be age related too. I had lived some 30-plus years before the Web became a thing ... and I kind of liked the 70's and 80's.
I feel the points made in this article have been debated many times on HN. The argument about power usage for developers, as opposed to consumers, is probably insignificant compared to the inefficient use of compute resources for deployed software today. Arguably LLMs are probably enabling some web developers to create native applications rather than Electron monstrosities saving many P-core CPU cycles multiplied by…
Does any such project exist, or are developers using LLMs to help them develop new Electron monstrosities? I think that if a developer has the sensibility to want to develop a native app, they will do so regardless, it's not that much harder.
> Models require massive amounts of electricity and water (to cool servers). Does it? https://prospect.org/environment/2024-09-27-water-not-the-pr... > training GPT-3 used as much water as just under twice what is required for the average American’s beef consumption. In other words, just two people swearing off beef would more than compensate. https://andymasley.substack.com/p/a-cheat-sheet-for-conversa... > It would…
> Models require massive amounts of electricity and water (to cool servers). Does it? https://prospect.org/environment/2024-09-27-water-not-the-pr... > training GPT-3 used as much water as just under twice what is required for the average American’s beef consumption. In other words, just two people swearing off beef would more than compensate. https://andymasley.substack.com/p/a-cheat-sheet-for-conversa... > It would…
The whole issue of "using" water is meaningless to me in the context of the water cycle. Does a data center "use" water? Whatever water evaporates from their cooling systems falls again as rain and becomes someone else's water. Same with farming - it all either evaporates (sometimes frustratingly right from the field it was applied to, or otherwise from the surface of the river it eventually runs into, or from the food you bite into, or from your sweat, or from your excretions/the sewage system/rivers again), or ends up supplementing a (typically badly depleted) aquifer, or gets temporarily used by animals including humans to e.g. hydrolyze fats (but full completion of metabolism of said fats & fatty acids actually returns MORE water on a net basis than it took to metabolize them) and so on.
In short, water is never passing into anything or anyone. It's passing through it. You don't own it, you're just borrowing it.
Even water recirculated as a coolant in a data center (the closest thing to actually "using" water) is a finite quantity, needed only one time, with maybe small top-ups due to losses, all of which end up, you guessed it, evaporating into the commons.
> “If you’re going to use generative tools powered by large language models, don’t pretend you don’t know how your sausage is made.” This isn't a very hopeful quote given how many people continue to eat sausages even though we all know how sausages are made.
and why focus on LLMs... there is a shitton of other things that use power/resources/... if we are going to be start worrying here we should consistently apply this across everything...
> Models require massive amounts of electricity and water (to cool servers). Does it? https://prospect.org/environment/2024-09-27-water-not-the-pr... > training GPT-3 used as much water as just under twice what is required for the average American’s beef consumption. In other words, just two people swearing off beef would more than compensate. https://andymasley.substack.com/p/a-cheat-sheet-for-conversa... > It would…
I must be missing something because water seems like the least of our worries when compared to the energy being used and the carbon dioxide being produced. I know I know, still talking about CO2 at parties instead of the hot new water thing? BOOOO-RINGGGG!!! The whole issue of "using" water is meaningless to me in the context of the water cycle. Does a data center "use" water? Whatever water evaporates from their coo…
Some places rely on deep aquifers that don't refill on human lifetimes, essentially fossil water. But that's mostly a local problem and we should just stop building water-hungry industry and agriculture in those places because it's stupid.
> Models require massive amounts of electricity and water (to cool servers). Does it? https://prospect.org/environment/2024-09-27-water-not-the-pr... > training GPT-3 used as much water as just under twice what is required for the average American’s beef consumption. In other words, just two people swearing off beef would more than compensate. https://andymasley.substack.com/p/a-cheat-sheet-for-conversa... > It would…
I don't understand the "whatboutism" angle. Running LLMs requires massive amounts of electricity and water. Period. That beef consumption and other activities also require massive amounts of resources is independent. Both can be true. And both need to be addressed. We no longer have the luxury to pick and choose which climate interventions we find most convenient; we need to succeed on all fronts. If the patient is d…
You use a laptop, you probably play some video games, you see movies, you probably eat meat, you probably drive a car around town instead of a velomobile.
So the question is, how much energy is used by AI and what does it get us? Google's latest TPUs consume ~250W per fp8-petaflop at the data-center level (ie including cooling etc, according to their recent presentations). Even assuming 50% utilization which is a bit poor, that's say 0.5kW per fp8-petaflop, or ~1kW per 10^15 dense-parameter-equivalent tokens per second. So using a huge model, say 10T dense-equivalent parameters, and doing inference at the rate of o3-mini or Gemini 2.5 Pro (around 150 tokens a second) consumes 1.5kW during generation time. But maybe a better way to think about it is just that you'd be using 10J of energy per token, or ~15J of energy per word.
In that context then, generating a book the length of Fellowship of the Ring (~180k words) would consume ~0.75kWh of energy. That's about like playing a video game on a PS5 for ~3 hours. In the US, that's on average ~270g of CO2, like driving ~1.5 miles or so in a Prius. It'd be about like riding ~30 miles on an e-bike.
Another way to think of it might be that if you are texting with a friend, wherein lets say you both aggressively type at ~120wpm, you'd be using ~15W of power to chat with an AI at a similar pace, about the power draw of your friend's Macbook Air he would be typing on. That's around 0.5-0.7 miles on an ebike per hour of chatting.
So even compared to your typical leisure activities, chatting with an AI is comparable in resource use to whatever else you'd be doing in that time most likely. And of course, a human in the US on average generates ~1.5kg of CO2 per hour they're alive in the US. In Bolivia, it's ~150g/hr, and in Sudan it's ~50g/hr or so. So if you're a company and you want to be climate conscious merely hiring someone who is a ~ subsistence farmer to do a job such that their standard of living rises enough to roughly emit the CO2 of the average Bolivian would mean they emit more carbon in a typical day than running an AI to produce as many words as are in the entire Game of Thrones book series to date. And if you were going to help someone immigrate from Bolivia to the US, you could have the AI write ~10 Songs of Ice and Fire a day for the same net CO2 output.
Not to say that we shouldn't do those things because of climate or whatever, but I'm just saying that the objective energy and climate impact of using these models for things is small compared to basically any alternative for completing a task and most entertainment activities people do (movie theaters for instance use ~7kW or so while running, so in the time you spend watching the Fellowship of the Ring in a theater, an AI could write a text as long as FOTR 28 times).
> “If you’re going to use generative tools powered by large language models, don’t pretend you don’t know how your sausage is made.” This isn't a very hopeful quote given how many people continue to eat sausages even though we all know how sausages are made.
> we all know how sausages are made. I'm doubtful. There's a few documentaries showing the process and I've had people tell me multiple times that they were genuinely shocked by it. I'm assuming those people "knew" and it's just that knowing "it's all the waste at the butcher getting stuffed into guts" is not quite the same as seeing it first hand.
Natural sausage casings are specialty items. If you're buying it at the grocery store, it's probably collagen (closely related to gelatin).
And it's not "all the waste". It includes fatty cuts that people wouldn't want to eat whole, but it doesn't include organ meats outside of specialty items.
Perhaps people find meat-grinding distressing, though it's really not all that different from ground beef. The emulsified filling of hot dogs and bologna looks odd, but the ingredients are inoffensive.
I'm less disturbed by sausage-making than by the slaughter and prime butchering of animals. They're no less dead and dismembered if you're eating a steak or pot roast. I'd rather we at least make use of all of the other parts.
Earlier quoted context omitted.
I don't understand the "whatboutism" angle. Running LLMs requires massive amounts of electricity and water. Period. That beef consumption and other activities also require massive amounts of resources is independent. Both can be true. And both need to be addressed. We no longer have the luxury to pick and choose which climate interventions we find most convenient; we need to succeed on all fronts. If the patient is d…
According to https://www.theverge.com/24066646/ai-electricity-energy-watt... a reasonable estimate is half a percent of global energy usage by 2027. On the one hand, this is a lot of energy. On the other, that's in the same general neighborhood as what we could save by maintaining proper tire pressure on our cars. (See https://www.factcheck.org/2008/08/the-truth-about-tire-press... for more about that.)