> By contrast, the unglamorous and largely disregarded business of making cement accounts for around 7 per cent of global emissions.
Oh, that's not a good example of the point they're trying to make. The emissions from concrete are a point of major concern and are frequently discussed. A ton of effort is being put into trying to reduce the problem, and there are widespread calls to reduce the use of the material as much as possible.
Many "new" expenditures replace existing stuff. The initial versions are often the worst iterations we'll see so even though the capability is going up, the energy usage will go down over time. It isn't universal (as we've seen a lot of new true growth), but it is common.
And what about the predictions of energy use that did pan out, like air conditioning and stuff? Also in 1999 how many personal computer companies were restarting nuclear power plants to fuel their projected energy consumption? Feels like a weird argument to make when the investments into AI I fra are literally measured in gigawatts. Feels like a weird argument in general - ai consuming lots of energy isn't some weird degrowth conspiracy theory
> But we have been here before. Predictions of this kind have been made ever since the emergence of the Internet
I don't think I live in the same world as the author. Ever since the emergence of the Internet, "stuff related to IT" has been using more and more energy.
It's like saying "5G won't use as much electricity as we are told! In fact 5G is more efficient than 4G". Yep, except that 5G enables us to use a lot more of it, and therefore we use more electricity.
Hopefully the panic continues and we get a lot of extra electricity, ideally via nuclear, wind, solar - and then if AI is a flop at least we get big progress on global warming.
“You may not know about the issue but I bet you reckon something, so why not tell us what you reckon. Let us enjoy the full majesty of your uninformed ad-hoc reckon” - David Mitchell.
> But far from demanding more electricity personal computers have become more efficient with laptops mostly replacing large standalone boxes, and software improvements reducing waste.
If only it was true, I reckon we’re using multiple-orders of magnitude more computational per $ of business objectives simply because of the crazy abstractions. For example, I know of multiple small HFT firms that are crypto market makers with their trading bots in Python. Many banks in my country have excel macros on top of SQL extensions on top of COBOL. We’ve not reduced waste in software but rather quite the opposite.
I don’t think this is super relevant to the articles point but I think it’s an under discussed topic.
AI helped me fix my own car, no new parts, no driving to the stealership, no comfy lobby to light, no extra building to heat, no IT system to book me into...
It's my opinion AI, like many technologies since the 1950s, will lead to more dematerialization of the economy meaning it will net net save electricity and be "greener".
This is an extension of what steven pinker says in Enlightenment now.
And that's just an example, there are many power-related deals of similar magnitude.
The companies building out capacity certainly believe that AI is going to use as much power as we are told. We are told this not on the basis of hypothetical speculation, but on the basis of billions of real dollars being spent on real power capacity for real data centers by real people who'd really rather keep the money in question. Previous hypotheses not backed by billions of dollars are not comparable predictions.
1. Microsoft is not a "dotcom startup". We live in a different, much more consolidated tech world of companies that either balloon to become behemoths or are bought by existing behemoths.
2. Power and data centers can be used for other things than AI.
3. They might turn out to be wrong and not need the deal/power. For companies sitting on a shitload of cash that would be an inconvenience whereas not investing and then later having to beg for electricity amounts to losing the race.
That's an awful amount of certainty for something that isn't backed by very much certainty at all. Just "previous claims about inefficiency in tech have ended up being incorrect".
As a counterpoint: look at crypto. The amount of power used by cryptocurrency has _not_ gone down, in fact it's increased.
There are contrary trends: LLM’s are getting lots of efficiency improvements, but they’re being used more.
Which is more important? Understanding what happened so far is impossible without data, and those trends can change. It depends on what new technologies people invent, and there are lots of smart researchers out there.
Armchair reasoning isn’t going tell us which trend is more important in the long term. We can imagine scenarios, but we shouldn’t be very confident about such predictions, and distrust other people’s confidence.
> Most of the increase could be fully offset if the world put an end to the incredible waste of electricity on cryptocurrency mining (currently 0.5 to 1 per cent of total world electricity consumption, and not normally counted in estimates of IT use).
I do not accept this. It was once true under Proof-of-Work (typically ~1,000–2,000 kWh per transaction), not so much under Proof-of-Stake (typically 0.03–0.05 kWh per transaction).
Note that proof-of-stake may actually have a lower energy footprint than credit card or fiat banking transactions. An IMF analysis [1] pegged core processing for credit card companies at ~0.04 kWh per transaction (based on data centers and settlement systems), but noted that including user payment means like physical cards and terminals could increase this by about two orders of magnitude—though even then, it doesn't extend to bank branches or employee overhead - an overhead not implicit in decentralized finance.
Datacenters didn't need water cooling before the AI explosion. (Air cooling was still possible)
At first, DW's estimate was one drop of potable water was consumed for each query (normal queries, not more expensive ones)
The Google, I don't know who allowed the sincerity, God bless him, released a first hand analysis of their water consumption, and it is higher that the one drop estimate: 5 drops
OpenAI yesterday announced[1] a partnership to deploy computer chips, but chose to denominate the size of the deal in gigawatts (instead of dollars, or some measure of computing capacity, or some measure of capability). They certainly seem to think about this in terms of electricity requirements, and seem to think they require a lot of it.
(I may have the units off a bit, but it looks like OpenAI's recent announcement would consume a bit more than the total residential electricity usage of Seattle.)
What’s the energy profile of running inference in a typical ChatGPT prompt compared to:
- doing a google search and loading a linked webpage
- taking a photo with your smartphone and uploading it to social media for sharing
- playing Fortnite for 20 minutes
- hosting a Zoom conference with 15 people
- sending an email to a hundred colleagues
I’d be curious. AI inference is massively centralised, so of course the data centres will be using a lot of energy, but less centralised use cases may be less power efficient from a wholistic perspective.
I found this video https://youtu.be/IQvREfKsVXM interesting, especially because it mentions couple of AI studies/papers that argue in favor of much smaller (and more efficient) models. (And I have never heard of them.)
I suspect that yes, for AGI much smaller models will eventually prove to be sufficient. I think in 20 years everyone will have an AI agent in their phone, busily exchanging helpful information with other AI agents of people who you trust.
I think the biggest problem with tech companies is they effectively enclosed and privatized the social graph. I think it should be public, i.e. one shouldn't have to go through a 3rd party to make an inquiry for how much someone trusts a given source of information, or where the given piece of information originated. (There is more to be written about that topic but it's only marginally related to AI.)
I'd be very interested in seeing some kind of aggregated daily demand curve for AI workloads.
It seems like a lot of the hyperbolic angles are looking at this as a constant draw of power over time. There is no reason for a GPU inference farm to be ramped up to 100% clock speed when all of its users are in bed. The 5700XT in my computer is probably pulling a mere 8~12W right now since it is just sitting on an idle desktop. A hyperscaler could easily power down entire racks based upon anticipated demand and turn that into 0W.
38 comments
[ 3.6 ms ] story [ 53.2 ms ] threadOh, that's not a good example of the point they're trying to make. The emissions from concrete are a point of major concern and are frequently discussed. A ton of effort is being put into trying to reduce the problem, and there are widespread calls to reduce the use of the material as much as possible.
I don't think I live in the same world as the author. Ever since the emergence of the Internet, "stuff related to IT" has been using more and more energy.
It's like saying "5G won't use as much electricity as we are told! In fact 5G is more efficient than 4G". Yep, except that 5G enables us to use a lot more of it, and therefore we use more electricity.
It's called the rebound effect.
If only it was true, I reckon we’re using multiple-orders of magnitude more computational per $ of business objectives simply because of the crazy abstractions. For example, I know of multiple small HFT firms that are crypto market makers with their trading bots in Python. Many banks in my country have excel macros on top of SQL extensions on top of COBOL. We’ve not reduced waste in software but rather quite the opposite.
I don’t think this is super relevant to the articles point but I think it’s an under discussed topic.
Indeed. But that is because we optimized (and are still optimizing) for speed of development, not much else.
It's my opinion AI, like many technologies since the 1950s, will lead to more dematerialization of the economy meaning it will net net save electricity and be "greener".
This is an extension of what steven pinker says in Enlightenment now.
And that's just an example, there are many power-related deals of similar magnitude.
The companies building out capacity certainly believe that AI is going to use as much power as we are told. We are told this not on the basis of hypothetical speculation, but on the basis of billions of real dollars being spent on real power capacity for real data centers by real people who'd really rather keep the money in question. Previous hypotheses not backed by billions of dollars are not comparable predictions.
2. Power and data centers can be used for other things than AI.
3. They might turn out to be wrong and not need the deal/power. For companies sitting on a shitload of cash that would be an inconvenience whereas not investing and then later having to beg for electricity amounts to losing the race.
As a counterpoint: look at crypto. The amount of power used by cryptocurrency has _not_ gone down, in fact it's increased.
Which is more important? Understanding what happened so far is impossible without data, and those trends can change. It depends on what new technologies people invent, and there are lots of smart researchers out there.
Armchair reasoning isn’t going tell us which trend is more important in the long term. We can imagine scenarios, but we shouldn’t be very confident about such predictions, and distrust other people’s confidence.
I do not accept this. It was once true under Proof-of-Work (typically ~1,000–2,000 kWh per transaction), not so much under Proof-of-Stake (typically 0.03–0.05 kWh per transaction).
Note that proof-of-stake may actually have a lower energy footprint than credit card or fiat banking transactions. An IMF analysis [1] pegged core processing for credit card companies at ~0.04 kWh per transaction (based on data centers and settlement systems), but noted that including user payment means like physical cards and terminals could increase this by about two orders of magnitude—though even then, it doesn't extend to bank branches or employee overhead - an overhead not implicit in decentralized finance.
[1] https://www.elibrary.imf.org/view/journals/063/2022/006/arti...
At first, DW's estimate was one drop of potable water was consumed for each query (normal queries, not more expensive ones)
The Google, I don't know who allowed the sincerity, God bless him, released a first hand analysis of their water consumption, and it is higher that the one drop estimate: 5 drops
https://services.google.com/fh/files/misc/measuring_the_envi...
(I may have the units off a bit, but it looks like OpenAI's recent announcement would consume a bit more than the total residential electricity usage of Seattle.)
1 - https://openai.com/index/openai-nvidia-systems-partnership/
I suspect that yes, for AGI much smaller models will eventually prove to be sufficient. I think in 20 years everyone will have an AI agent in their phone, busily exchanging helpful information with other AI agents of people who you trust.
I think the biggest problem with tech companies is they effectively enclosed and privatized the social graph. I think it should be public, i.e. one shouldn't have to go through a 3rd party to make an inquiry for how much someone trusts a given source of information, or where the given piece of information originated. (There is more to be written about that topic but it's only marginally related to AI.)
It seems like a lot of the hyperbolic angles are looking at this as a constant draw of power over time. There is no reason for a GPU inference farm to be ramped up to 100% clock speed when all of its users are in bed. The 5700XT in my computer is probably pulling a mere 8~12W right now since it is just sitting on an idle desktop. A hyperscaler could easily power down entire racks based upon anticipated demand and turn that into 0W.