It is rather unfortunate to have to explain this in a free software community. But "AI" (please use "LLMs" instead) is not the revolutionary technology you think it is, certainly nothing equivalent of electricity.
I say this as someone with a math background who studied the foundations of statistical learning behind the current wave of machine learning. The brute force methods using statistical inference have a theoretical upper limit given bounds on resources. The companies have unbashedly abused all resources possible for this. Never in human history have we put in so much resource on something that is _this_ experimental. All of the statistical foundations of ML were more or less settled in the 1990s, the reason they didn't take off back then is because of two fundamental blockers: lack of huge amounts of data to train on and lack of extensive amount of compute. Two things these statistical methods were extremely greedy of. I am not against machine learning at all, I am against methods that are inherently resource hungry, inefficient and have been weaponized against ecology, used for genocide and against human dignity of labour. Your claim that AI companies will start to be more efficient _when_ resources run out is absurd. Why should we trust them one bit when they have shown nothing but incompetence and frugality for so long? Either way, nobody knows how History moves. Right now it's been close to 5 years already and we have seen nothing but damage. We have put a total of 1.4 trillion (!) [0] dollars in this and only one major company is making profit out of it. That is more than 25% of my country's (India) GDP. Is this supposed experimental technology worth 25% of a big country's economy? Google's own engineers make memes about how LLMs are destroying their codebases and how they hate it [1]. Instagram's LLM started handing out login credentials to people out of nowhere [2]. And we might go into the never-ending debate of how these are only contingent problems and once we have "good enough models", these would go away. I am sorry, but that is exactly the ideology these companies would have you believe. Again, these methods are mathematically doomed to be resource hungry. I would be happy to support, talk and experiment with other methods to machine intelligence, especially ones that are grounded in formal verification and symbolic methods than purely stochastic ones. (And no, just having a formal prover as a guarding rail for LLMs is not good, it's still _very_ resource intensive and inefficient.) To me GCD008 is not a loaded gun at our feet. It's us trying to finally take a stand for our dearest activity: programming, in an age where that is being destroyed by several entities. Programming is and shall remain a human-centered activity, or to quote from our favorite SICP: > Programs must be written for people to read, and only incidentally for > machines to execute. Just like mathematics is not about churning out proofs but the very activity of using mathematical tools to reason through problems, programming is not equivalent to code generation. So, if we are to do a proper historiography of technology then I'd request us to be more rigorous. The equivalence being made that "AI" is in a problematic stage right now and will improve later on because of shortages (caused by it's own inefficiency), what evidence do we have of that? Meta burned 80 billion dollars in building virtual worlds that it also thought would get better with time as VR technology catches on, it didn't. Your statement about equivalence is faulty because every technology that failed used exactly this rhetoric until it died. This is exactly what the 60s-70s era of "AI" (the symbolic camp) believed. This is the "AI" our folks Stallman, Sussman and original Lispers were from. And last I checked, Sussman is disappointed that at least they attempted and failed with a better approach that wasn't this resource hungry [3]. But the point is, "AI" replacing everything was the peak ideology during 70s, and if they had so many intelligent computer scientists with them (McCarthy, Stallman, Sussman, Knuth, Greenspun, Raj Reddy, et.al) and they couldn't have predicted the failure early on, then what makes us be more optimistic? We have more intelligent researchers? Our methods are inherently superior? The only thing that makes us have confidence is that we live in 2026, a globalized capitalist economy where they can outsource scrap work like data labelling from workers in my country and its neighbors. They now have an unimaginable amount of compute and have the capacity to block 40% of the entire world's RAM at their will. They have more stolen data than ever from us and our family members, and they sell it for their vicious practices. And supposedly all of this bad and ugly stuff is because they will in a few years after burning a few trillions will give us some revolutionary technology? That will correct all of these wrongs? That will correct the breached privacy? The torture that they're now doing to artists? Can we really, in good conscience, put all of these terribe things under the rug because this is supposedly a disruptive technology? I honestly can't do this in good conscience, without simply having to buy into their ideology. I would be happy to continue with this historiography, but I think I have already said more than enough. A good reference I'd like to provide that situates similar criticisms is Piero Scaruffi's _Intelligence is Not Artificial_ (2015). It doesn't involve that much math, but enough to make the argument for why deep learning has theoretical inefficiencies within it. I have a few other books similar to this one if people are interested in doing a proper historiography of this technology than what meets our eye immediately. [0]: https://isaiprofitable.com/ [1]: https://www.404media.co/google-employees-internally-share-memes-about-how-its-ai-sucks/ [2]: https://www.404media.co/hackers-simply-asked-meta-ai-to-give-them-access-to-high-profile-instagram-accounts-it-worked/ [3]: https://dustycloud.org/blog/sussman-on-ai/ Regards, On 2 June 2026 11:43:41 GMT+05:30, Pjotr Prins <[email protected]> wrote: >Dear all, > >I think the environmental statement is the weakest part of GCD008, or >at least the most exaggerated. There is no denying that *current* AI >deployments are energy intensive and dirty - just like coal was in the >19th century driving the industry revolution - but that is simply due >to stupidity and, in part, to the early days of the success of >transformers and based on the simplistic idea that bigger data centers >lead to better models. This is already proving not to be true because >of the flattening out of model performance. And whatever way you look >at it, it is peanuts compared with other emissions. In the larger >sense, CO2-wise we live in the age of crazy. > >The world certainly has bigger fish to fry. > >Even the statement that we are running out of chips is a red herring. >Either AI companies get smarter or the vendors get smarter. Any >shortage will drive that. > >We simply have to get smarter about models. Economics will drive >that and an energy crisis helps (in that way). Having 8+ >competing initiatives world wide does not help either, or maybe it >does :) > >Ludo, do you realise this may be a good thing? The AI peeps will have >an incredible incentive to become more energy efficient (when the >money starts running out) and that may lead to better energy >technology for all of us? Good things can happen because of bad >things. Huge investments, such as the terrible 2nd WW, can lead to >incredible progress - including jet engines and computers ;). >Even Musks' (probably stupid) idea of shooting data centers into >space could be one. It is rather likely that building models will run out >of steam, and inference can run on your laptop. Give it 5 years. > >Automating package writing in Guix is a boon for *our* projects. It *is* >saving us significant time with better outputs. I get things done I >could not do before, both in terms of my scientific effort and my >software engineering effort. I need to use AI for my work. There is no >other way (any more). We are planning a massive refactoring of >GeneNetwork - we would not have considered before. There are projects >in Guix we have not touched that could benefit. I would love to use >true open models and create models deterministically with free software >instead of the industry tooling, so we get back our transparancy and >control. That is where our effort should be going. Our effort should >not go to knee jerk statements about why AI is bad for us. Guix is >sending the wrong message and educating people in the wrong way. > >What I don't like about GCD008 - and I'll say it again - is that it >makes us look like dinosaurs and it splits our community. It won't be >the end of Guix, but the unintended side effect is that we are keeping >everything small while the world changes around us. We are part of >*change* and we should make it abide by our ideas of controlling our >own destiny. I don't mind a policy, per se, or even a pledge, but I do mind >being made to look like an idiot. And that from someone who's >ancestors where partial to famous church splits ;). You know I like >free software and minimalism, but I don't like GCD008 because it is a >loaded gun aimed at our own feet. > >Rejecting AI, in my book, is like rejecting electricity or medicine. >Sure, we can do without electricity or medicine. But not many would >want to. Today. > >Pj. > >On Mon, Jun 01, 2026 at 09:35:31PM +0200, Ludovic Courtès wrote: >> Hi Greg and all, >> >> Greg Hogan <[email protected]> skribis: >> >> >> > One >> >> > expressed concern for generative AI is over datacenter energy and >> >> > water use. The Guix project currently employs multiple large compute >> >> > clusters running tens of thousands of builds. >> >> >> >> Are you comparing ~50 GPU-less nodes worldwide (that I know of, but >> >> let’s imagine there are big secret Guix build farms and the figure is >> >> 1,000 nodes) to Colossus¹ & co.?… >> >> [...] >> >> > The Colossus-scale models have billions of global users. Guix has >> > perhaps a few thousand? The analysis would be a speculative comparison >> > of the per-capita energy use of the Guix project with and without the >> > use of LLMs. >> >> The carbon footprint of Guix build farms (meaning the greenhouse gas >> emissions to produce the hardware of those build farms and to run them) >> is proportional not to the number of users but to the number of >> packages. >> >> If Guix had a billion users, with a comparable number of packages, the >> build farm size would comparable to what we have today. We would need >> to use a content delivery network (CDN), for example, which adds to the >> carbon budget of the infrastructure, but relatively little. >> >> The carbon footprint of facebook.com, with its billion users, is >> certainly very high, but its infra is a toy compared to what companies >> like OpenAI, xAI, and Anthropic are building. The popularity rise of >> Facebook or even YouTube etc. did not lead to a shortage of RAM and GPUs >> on the market contrary to what we’re experiencing today. >> >> You may say: “with AI, I might do in 1 day what I would otherwise do in >> 10 days, proof that it’s more efficient and thus reduces my carbon >> footprint”. That’s not the case due to Jevon’s paradox, as Simon nicely >> explained: >> <https://codeberg.org/guix/guix-consensus-documents/pulls/13#issuecomment-15843803>. >> >> In the 2015 Paris Agreement¹, countries pledged to act to keep global >> surface temperature increase below 1.5℃ compared to pre-industrial era. >> To achieve that we need to significantly reduce GHG emissions (for >> Western Europe where I live, we’re talking about going from ~10 TCO₂e to >> 2 TCO₂e per capita—this is huge). “AI” takes us in the opposite >> direction, and fast. >> >> >> *** Why are we discussing all this in the context of Guix? *** >> >> The point I am trying to make in GCD 008 on this topic is that free >> software doesn’t exist in a vacuum. Yes, LLMs (+ agents) can automate >> away tedious packaging tasks. But we can acknowledge that the >> environmental cost of these tools *alone* is a problem for society, and >> we can refuse *collectively* to not make ourselves dependent on it. >> >> After all, we’ve been running this project without LLMs for 14 years, so >> it shouldn’t be too hard, right? Yes, we have work to do to improve >> package maintenance automation; luckily, we already have solid >> foundations for that—more solid than most other distros. >> >> I emphasized “collectively” above: the GCD has never been about policing >> people—as individuals, we have to make do and sometimes compromise on >> what we think is “right”, from running non-free software to using LLMs; >> that’s okay and explicitly acknowledged in the document. >> >> I hope I sufficiently clarified and contextualized what I had in mind >> regarding this point. >> >> >> *** Is the GCD about “ecology”? *** >> >> No it’s not. The environmental cost of “AI” is an important point IMO >> but there are several more. >> >> The other points delve into social, ethical, and legal issues. Legal >> issues have been widely discussed already, but the rest much less. >> >> I would invite people to take a look at these other points and the >> references in the document. Summarizing it as “AI is bad” would not >> only be dismissive, it would miss out on issues that affect free >> software projects. >> >> Free software is a social endeavor with an ethical compass. Just like >> some were famously critical of the non-free software running in their >> printer or pacemaker and what it means for their agency, questioning how >> these technologies affect us and society seems totally legit. >> >> >> *** “My motivation is different!” *** >> >> That’s okay! The pledge in the document derives from my motivation. >> Perhaps with a different motivation you reach similar conclusions and we >> can come up with a pledge that addresses both our concerns. >> >> The “Motivation” section gives context, but what matters in the end is >> the pledge. >> >> >> Ludo’. >> >> ¹ https://en.wikipedia.org/wiki/Paris_Agreement >> > >-- > Divya Ranjan Pattanaik Mathematics, Philosophy & Libre Software
