HN reacts to a New Yorker piece on the “obscene energy demands of AI” with exactly the same arguments coiners use when confronted with the energy cost of blockchain - the product is valuable in of itself, demands for more energy will spur investment in energy generation, and what about the energy costs of painting oil on canvas, hmmmmmm???
Maybe it’s just my newness antennae needing calibrating, but I do feel the extreme energy requirements for what’s arguably just a frivolous toy is gonna cause AI boosters big problems, especially as energy demands ramp up in the US in the warmer months. Expect the narrative to adjust to counter it.
Exchange presented without comment:
My prediction: the advance of tech by AI will far surpasse what it consume in energy.
To look at the energy consumption of current model is extremely short sighted. If AI create a new material, a new solar cell, advance fusion reactor is all of humanity that jump forward.
Furthermore new generation of AI accelerators and new algorithms will improve efficiency by order of magnitute, it’s still early days.
For every good thing, come up with a bad.
The material created will be a better poison/virus. The algorithm to keep the fusion tokamak from going boom will be at best 99% correct. The new solar cell? More exotic materials required than the current.
Blind optimism is a vice we cannot afford.
The post you’re responding to doesn’t argue from blind optimism, it argued a reasonably-expected gain in net beneficial effects.
Amazing how they just assume that this new future AI will actually work. Like there were not several AI winters.
or that the existing AI works. Some of the worst social damage is from AI systems deployed in the wild that don’t even do their fucking job properly in the first place, let alone doing it evilly. (Many deployments are both evil and incompetent.)
Yeah and this damage keeps happening, see just how much damage faulty IT does when we could fully understand the code and processes behind it. For example https://en.wikipedia.org/wiki/British_Post_Office_scandal (at least 4 suicides!). AI is only going to make it worse.
Jesus fucking christ, if COLD FUSION is a REASONABLY EXPECTED GAIN then what on earth would an UNREASONABLE gain be? Infinite blowjobs? Quality discourse on HN?
Your “frivolous toy” argument interested me, apparently video games use 34 TWh a year
If the random dude on HN is to be believed chatGPT uses 182M kWh or 0.182TWh a year.
Not sure what to make of that.
christmas lights bring people actual joy re
again, this is literally a reskinned bitcoiner argument
More👏power👏for👏vibrators👏
Almost no one in here would complain about more science spending, but would complain about capitalism (me too capitalism bad). What is the utility of going to space? It requires lots of energy but it’s still valuable. What about studying random new species in remote locations? The mating habits of microorganisms? And the common refrain is you can’t always know what the next discovery will be so sometimes you must be a little directionless in your approach because you might not even know where to look in the first place.
AI research is science, it’s academic. The problem is it’s been entirely co-opted by capitalist corporations, but that’s besides the point. AI research isn’t inherently not worth doing because a bad organization is trying to do it.
The ideal solution would be for the research money to come from governments and given to academic or other public institutions who will do their science in the open instead of in closed labs trying to monetize everything.
So I agree with your premise in general but it’s slightly more nuanced than that. Probably LLMs or stable diffusion used on a mass scale is little more than a toy and not worth the energy costs, but 1. It will get better, and in the case of language models, that could have profound impacts on society and 2. There are other kinds of AI use cases, such as alphafold and the materials research Deepmind published 3. There are other things being researched that are equally important that get little daylight, such as robotics and agentic AIs
non sequitur
There are other kinds of AI use cases, such as alphafold and the materials research Deepmind published
alphafold is mostly pattern matching on known proteins and the other bit, well google very quickly distanced themselves from these shitty results when they learned how shitty they are. i’ve made a post about it specifically https://discuss.tchncs.de/post/11138402 and i won’t rewrite it again
It is in fact quite doable to condemn the AI industry, which is what uses all the energy, without condemning AI scholarship, which quite definitely doesn’t. In fact, I know that’s possible, because that’s the overall take of the commentariat here.
The frivolous toy and the thing using all the energy are the same picture.
See also: supporting space research is why many of us condemn starlink’s LEO pollution.
I don’t disagree with anything you said but I don’t think the distinction was explicit in the OP hence my comment. It was a “yes and”
Also responding to a general anti AI trend in public discourse that I think is short sighted and inaccurate rather than specifically to OP
assume we’re all fully up on the discourse and the players here
From your syntax I can divine that you are mad at me (or rather, my submission) but for the life of me I don’t understand why. Is it because I wrote “AI” instead of “the bad AI from BigTech using TWh to generate shitty hero images for blogs but not the good AI from the heroic researchers constructing our glorious future for the pure love of science” ?
If so, nothing would please me more than for the “bad AI” to crash and burn, pauperize Sam Altman and all his bootlickers, and AI research to retreat to the academic caves to hibernate another AI winter.
every time I see responses like that I’m left wondering if it’s from someone working in (or closely adjacent to) the field. someone with some eyes on the potential and Big Mad about the bullshit, but feeling unable to affect it in any manner
a charitable interpretation (and wishful thinking perhaps?), but still
I’m not mad at you at all, i basically agreed with you. i was just engaging in a more nuanced discussion…. Or trying to. the replies here seem fairly hostile so I think I’ll see myself out of this community.
ah yes, the type of nuance that can’t survive even the extremely mild amount of pushback you’ve experienced in this thread. but since we’re “fairly hostile” and all that, how about I make sure your lying AI-pushing ass can’t show up in any of our threads again
I should’ve known taking my time to explain our stance was a waste of my fucking time when you brought up nuance in the first place — the only time I see you shitheads give a fuck about that is when you’re looking to shift the Overton window while pretending to take a centrist position
this is the general argument in favour of cryptocurrency, with the name changed. you don’t seem to have argued that the actual reality of AI we have right now is not the same problem.
Because I’m not arguing with OP, I’m largely agreeing with them. Generating silly images and doing school kids homework is not the promised land of AI the corporate overlords keep promising. But that’s not to suggest the field in general has zero uses. Crypto and AI are apples and oranges and while I’m not exactly sure what you mean by the arguments being the same, it would be possible for the same argument to be true for AI and not true for crypto, because AI has much more obvious use cases to benefit the common good.
“AI” is a marketing term for various at best slightly related technologies. If you mean LLMs or whatever, you’d need to be specific else you’re not even defining the goalposts before setting them up with wheels.
yeah, I definitely think machine learning has obvious use cases to benefit the common good (youtube auto captions being Actually Pretty Decent Now is one that comes to mind easily) but I’m much less certain about most of the stuff being presently marketed as “AI”
i’m pretty cool with ELIZA
Can you tell me more about why you’re pretty cool with ELIZA? 😉
Is it that you would like to be able to tell you more about why you’re pretty cool with ELIZA?
we’re talking about you not me. come come elucidate your thoughts. can you elaborate on that?
Exactly. Some machine learning is great (image recognition for accessibility, machine translation) and some machine learning is awful (image recognition for cops, or machine translation for cops). But AI®️™️ is just mouth noises.
(also obviously image recognition and machine translation are at least real, even when for cops, as opposed to “creative thought produced by LLMs” which is even worse than mouth noises.)
AI is the name of the field of study. It has existed since the 60s. LLMs are neural networks one of the first and most widely used forms of AI.
how come the reply humans from programming dot dev have always the daftest takes?
who was this post for
It’s not even the right decade; the Dartmouth Summer Research Project on Artificial Intelligence was in 1956.
rubber duck replying, with a stuck posting key
I hear what you’re saying, but I think it’s sort of a motte-and-bailey setup:
Motte: Many functions can be probably approximately learned, even some uncomputable functions
Bailey: Consciousness, appreciation for art, useful laboring, and careful argumentation are learnable functions
- It will get better, and in the case of language models, that could have profound impacts on society
why is that a given?
the materials research Deepmind published
these results were extremely flawed and disappointing, in a way that’s highly reminiscent of the Bell Labs replication crisis
- There are other things being researched that are equally important that get little daylight, such as robotics and agentic AIs
these get brought up a lot in marketing, but the academic results of attempting to apply LLMs and generative AI to these fields have also been extremely disappointing
if you’re here seeking nuance, I encourage you to learn more about the history of academic fraud that occurred during the first AI boom and led directly to the AI winter. the tragedy of AI as a field is that all of the obvious fraud is and was treated with the same respect as the occasional truly useful computational technique
I also encourage you to learn more about the Rationalist cult that steers a lot of decisions around AI (and especially AI with an AGI end goal) research. the communities on this instance have a long history of sneering at the Rationalists who would (years later) go on to become key researchers at essentially every large AI company, and that history has shaped the language we use. the podcast Behind the Bastards has a couple of episodes about the Rationalist cult and its relationship with AI research, and Robert Evans definitely does a better job describing it than I can
The pro-AI comments here remind me of possibly the worst opinion I’ve heard about fossil fuels, which was that we should burn them all now as fast as possible to develop newer technologies to magically solve all the problems generated from burning all the fossil fuels.
There are a few bitcoiners out there who try to claim, with a straight face, that more bitcoin mining = more greener.
As far as I can tell the argument basically boils down to “we’ll just use SO MUCH electricity that the utility will have no choice but to invest in green energy just to keep up” – but honestly I can’t make heads or tails of it.
Oh that’s accelerationist drivel if I’ve ever seen it.
Also, this has been bouncing around in my head ever since I’d learned of the term: “e/acc? More like, “Eek! Ack!”
Yea kinda like during a famine the best course of action is for everyone to hoard and binge on as much food as possible. That will force the crops to become more bountiful.
As far as I can tell the argument basically boils down to “we’ll just use SO MUCH electricity that the utility will have no choice but to invest in green energy just to keep up”
yep. and I like using the rule 34 defense on them: “if that were desired, someone already would’ve made it”
these fucking clowns permanently don’t want to acknowledge (or just stay willfully ignorant about?) the fact that it’s easier to do cost and regulatory arbitrage by hunting for presently-favourable miner locations from which to burn electricity than it is to invest into (and possibly, likely even, invent!) whole-ass new green tech with sufficient output for their preferred ourobouros
That’s it. That’s the argument.
