> As a user it's a great free ride though. Maybe there IS such a thing as a free lunch after all!
if you consider the massive environmental harm AI has and continues to cause, the people whose work has been stolen to create it, the impacts on workers and salaries, and the abuses AI enables that free lunch starts looking more expensive.
> the people whose work has been stolen to create it
"Stolen" is kind of a loaded word. It implies the content was for sale and was taken without payment. I don't think anyone would accuse a person of stealing if they purchased GRRM's books, studied the prose and then used the knowledge they gained from studying to write a fanfic in the style of GRRM (or better yet, the final 2 books). What was stolen? "the prose style"? Seems too abstract. (yes, I know the counter argument is "but LLMs can do more quickly and at a much greater scale", and so forth)
I generally want less copyright, not more. I'm imagining a dystopian future where every article on the internet has an implicit huge legal contract you enter into like "you are allowed to read this article with your eyeballs only, possibly you are also allowed to copy/paste snippets with attribution, and I suppose you are allowed to parody it, but you aren't allowed to parody it with certain kinds of computer assistance such as feeding text into an LLM and asking it to mimic my style, and..."
AI has been trained on pirated material and that would be very different from someone buying books and reading them and learning from them. Right now it's still up to the courts what counts as infringing but at this point even Disney is accusing AI of violating their copyrights https://www.nytimes.com/2025/06/11/business/media/disney-uni...
Facts aren't copyrightable. Expression is. LLMs reproduce expression from the works they were trained on. The way they are being trained involves making an unlicensed reproduction of works. Both of those are pretty straightforwardly infringement of an exclusive right.
Establishing an affirmative defense that it's transformative fair use would hopefully be an uphill battle, given that it's commercial, using the whole work, and has a detrimental effect on the market for the work.
The still is not transformative but the model reproducing it is obviously transformative. Other general purpose tools can be used to infringe and yet are non-infringing as well.
If I watch a movie, then draw a near perfect likeness of the main character from my very good memory, put it on a tshirt and sell the t-shirt. That is grounds for violation of copyright if the source isn't yet in the public domain (not guaranteed but open to a lawsuit).
If I download all content from a website that has a use policy stating that all content is owned by that website and can't be resold. Then allow my users to query this downloaded data and receive a detailed summary of all related content, and sell that product. Perhaps this is a violation of the use policy.
All of this hasn't been properly tested in the courts yet.. large payments have already been made to Reddit to avoid this, likely because Reddit has the means to fight this in court.. my little blog though, fair game because I can't afford to engage.
For sure, it's rich people playing rules for thee not for me. What's interesting is we'll discover on which side of the can-afford-to-enforce-its-copyright boundary the likes of NYTimes fall.
That’s not “data laundering and hiding evidence of infringement” though.
You’re talking about overt infringement, the GP was talking about covert infringement. It’s difficult to see how something could be covert yet not transformative.
I think there is case to be made that AI companies are taking the content - providing people with a modified version of that content and not necessarily providing references to the original material.
Much of the content that is created by people is done so to generate revenue. They are denied that revenue when people don't go to their site. One might interpret that as theft. In the case of GRRM's books - I would assumed they were purchased and the author received the revenue from the sale.
I think you are missing some context. They were using Anna's Archive! They paid for nothing, downloaded in violation of copyright, and processed it. They violated US copyright law even before they actually ingested it!
Yes, there are ethical differences to an individual doing things by hand, and a corporation funded by billions of investor dollars doing an automated version of that thing at many orders of magnitude in scale.
Also, LLMs don’t just imitate style, they can be made to reproduce certain content near-verbatim in a way that would be a copyright violation if done by a human being.
You can excuse it away if you want with reduction ad absurdum arguments, but the impact is distinctly different, and calls for different parameters.
The silver lining on this very dark cloud is that it seems to have renewed interest in nuclear power, though that was inevitable with the coming climate crisis I suppose.
Haha, I would have thought the reckless cuts of DOGE or willingness of the current US administration to rely on AI for decision making would have driven home exactly why governments can't be trusted to manage nuclear.
It's just too dangerous to leave it in the hands of people who don't believe in science, and value money, power, and ideology more than anything else.
Its happening now, and there is nothing to stop it happening again in future.
Nuclear power plants do not achieve the fission rates required for an explosion, and the system failsafes are designed entirely with an absent crew in mind. Not sure why you think it’s somehow smarter or safer to let them control a hundred gigawatts of solar power but not nuclear power.
"Further, the NRC rejected past attempts to switch to a model based on the "hormesis theory" that Trump seemingly supports—which posits that some radiation exposure can be beneficial. The NRC found there was "insufficient evidence to justify any changes" that could endanger public health, Science reported.
One health researcher at the University of California, Irvine, Stephen Bondy, told Science that his 2023 review on the science of hormesis showed it is "still unsettled." His characterization of the executive order suggests that the NRC embracing that model "clearly places health hazards as of secondary importance relative to economic and business interests.""
The problem is not so much the carbon footprint of just using LLMs, but the impact from building and running massive data centers needed to train the models. Check out this report from the ITU:
> According to the latest edition of the report, electricity consumption by data centers — which power AI development and deployment, among other uses — increased by 12 per cent each year from 2017 to 2023, four times faster than global electricity growth.
> Four leading AI-focused companies alone saw their operational emissions increase in the reporting period by 150 per cent on average since 2020. This rise in energy that is either produced or purchased – known as Scope 1 and Scope 2 emissions – underscores the urgent need to manage AI's environmental impact.
> In total, the amount of greenhouse gas emissions reported by the 166 digital companies covered by the report contributed 0.8 per cent of all global energy-related emissions in 2023.
It's worth noting that it's not all gloom and doom. As the report optimistically notes:
> Renewable energy adoption: 23 companies operated on 100 per cent renewable energy in 2023, up from 16 in 2022.
> Scope 3 consideration: The number of companies publishing targets on indirect emissions from supply chains and product use rose from 73 to 110, showing increasing awareness of industry impacts.
So the environmental impact is alarming, but at least companies seem to be growing more conscientious about it. We must continue to hold tech companies accountable for their environmental impact, and keep pushing for renewable energy.
They’re restarting coal-fired power plants to run AI datacenters. I don’t know what your personal threshold is for “massive” environmental harm, but that meets mine.
It has been widely reported. Here is an example not of a coal-fired but diesel and gas mobile power sources. If you spend time looking you will have no trouble finding sources.
> Last week, the Environmental Protection Agency (EPA) issued a rule clarification allowing the use of some mobile gas and diesel power sources for data centers. In a statement accompanying the rule, EPA Administrator Lee Zeldin claimed that the Biden administration's focus on addressing climate change had hampered AI development.
> "The Trump administration is taking action to rectify the previous administration's actions to weaken the reliability of the electricity grid and our ability to maintain our leadership on artificial intelligence," Zeldin said. "This is the first, and certainly not the last step, and I look forward to continue working with artificial intelligence and data center companies and utilities to resolve any outstanding challenges and make the U.S. the AI capital of the world."
AWS is perhaps the most established vendor in this space. They if any should have their architecture sorted out and have great execution on their growth.
Without getting in to the fact that none of those people operate datacenters, what does that mean to you? You think Trump and Co are doing this as a purely political move to irritate environmentalists?
> In another move, DOE on Tuesday said it was offering loan guarantees for coal-fired power plant projects, such as upgrading energy infrastructure to restart operations or operate more efficiently or at a higher output.
Training AI models uses a large amount of energy (according to what I've read / headlines I've seen /etc), and increases water usage. [0] I don't have a lot to offer as proof, merely that this is an idea that I have encountered enough that I was suprised you hadn't heard of it. I did a very cursory bit of googling, so the quality + dodginess distribution is a bit wild, but there appear to be indiustry reports [2, page 20] that support this:
"""
[G]lobal data centre electricity use reached 415 TWh in 2024, or 1.5 per cent of global electricity consumption.... While these figures include all types of data centres, the growing subset of data centres focused on AI are particularly energy intensive. AI-focused data centres can consume as much electricity as aluminium smelters but are more geographically concentrated. The rapid expansion of AI is driving a significant surge in global electricity demand, posing new challenges for sustainability. Data centre electricity consumption has been growing at 12 per cent per year since 2017, outpacing total electricity consumption by a factor of four.
"""
The numbers are about data center power use in total, but AI seems to be one of the bigger driving forces behind that growth, so it seems plausible that there is some harm.
USA uses 21.3 TWh of petroleum per day for transportation. Even if AI was fully responsible for all data center usage (it is not even close) we're quibbling over 20 days of US transportation oil usage, which actually has devastating effects on the environment.
Data centers are already significant users of renewable electricity. They do not contaminate water in any appreciable amount.
There's an "AI is using all the water" meme online currently (especially on Bluesky, home of anti-AI scolds), which turns out to come from a study that counted hydroelectric power as using water.
As an example, Ren and his colleagues calculated the emissions from training a large language model, or LLM, at the scale of Meta’s Llama-3.1, an advanced open-weight LLM released by the owner of Facebook in July to compete with leading proprietary models like OpenAI's GPT-4. The study found that producing the electricity to train this model produced an air pollution equivalent of more than 10,000 round trips by car between Los Angeles and New York City.
(https://news.ucr.edu/articles/2024/12/09/ais-deadly-air-poll...)
> The study found that producing the electricity to train this model produced an air pollution equivalent of more than 10,000 round trips by car between Los Angeles and New York City.
I am totally on board with making sure data center energy usage is rational and aligned with climate policy, but "10k trips between LA and NY" doesn't seem like something that is just on its face outrageous to me.
Isn't the goal that these LLMs provide so much utility they're worth the cost? I think it's pretty plausible that efficiency gains from LLMs could add up to 10k cross USA trips worth of air pollution.
Of course this excludes the cost of actually running the model, which I suspect could be far higher
> 10,000 round trips by car between Los Angeles and New York City.
That seems like very low impact, especially considering training only happens once. I have to imagine that the ongoing cost of inference is the real energy sink.
It doesn't happen only once. It happened once, for one version of one model, but every model (and there are others much larger) has its own cost and that cost is repeated with each version as models are continuously being retrained
if you consider the massive environmental harm AI has and continues to cause, the people whose work has been stolen to create it, the impacts on workers and salaries, and the abuses AI enables that free lunch starts looking more expensive.