- cross-posted to:
- technology@lemmy.world
- cross-posted to:
- technology@lemmy.world
You must log in or register to comment.
Most of these arguments are conflating AI with the people using them.
Labor exploitation, scraping without consent, military contracts, surveillance policing, grid strain and hardware prices driven up by speculation are real harms. But none of them are properties of a statistical language model.
They come from concentrated capital and state power, and whatever technology arrives next will get used the same way.
Just like we’ve seen everything in your list, done by the same people, using different technology.
We have had hardware speculation during the crypto craze.
These same companies (Google) have been taking your data without consent to power their surveillance advertising.
Datacenters have always been power and water hungry, the ones powering LLMs are exactly the same ones that exactly the same companies have been building since the Internet began.
Essentially every major advance in technology has been used for state surveillance and in the military. AI is simply the latest capability.
Models reproducing licensed code, poor code quality and floods of junk contributions. These are real problems with the technology itself. And the solution to most of these things is to iterate, build better models trained on consentually obtained data.
Yes, you should still be mad at what the AI companies are doing. I am too. Just keep in mind that Google was evil before Gemini and will certainly be on the forefront of fuckery with whatever technology comes next as long as people keep getting tricked into misdirecting their anger onto the technology.
This is mostly true, but you’re glossing over the fact that AI will always be tied to concentrated capital due to the cost of training state of the art models. I understand why you argue that these flaws are not inherent to the technology, but in a practical sense, they might as well be considered inherent. It’s unlikely that the market will open up any time soon so that you can get your LLM from someone who isn’t a fascist billionaire tech bro.
This is mostly true, but you’re glossing over the fact that AI will always be tied to concentrated capital due to the cost of training state of the art models.
I think this is a false assumption.
We don’t know what the ultimate state of AI will be, we only know that we’re at the beginning of the journey.
Look at other technologies in similar positions. At one point computers were like this. Nobody but the military and the largest companies could afford them and now you can buy children’s toys with more storage and processing power than was available to the entire US military at the dawn of computers.
We had no idea what computers would become, even the experts making guesses that sounded wild at the time were off by several orders of magnitude. Bill Gates once said “640K [of RAM] ought to be enough for anybody”, modern PCs have around 100,000 times that.
The reason that AI seems like it’s only affordable to concentrated capital is because models using a large amount of parameter require a large amount of RAM. The 5TB of RAM estimated for some of the largest frontier models is a lot, about 100x a workstation PC. That’s a few orders of magnitude less than 100,000x. Given equal scaling speed, we’re closer to running frontier models on our desktop than we are to the dawn of computers.
In addition, the research is leaning towards there being a ceiling on how useful it is to add more parameters (use more RAM). Doubling parameters doesn’t result in doubling of capability so there is effectively a soft cap on the amount of parameters.
Hardware is growing near exponentially while the RAM demands of models have a ceiling.
Keep in mind that this only relates to the state of the largest LLMs. You can train generative models on consumer graphics cards. Object recognition, classification, etc are all tiny compared to LLMs. These are also AI and, I’d argue, that they actually produce more value than LLMs (which are failing to deliver on their promise of replacing white collar labor).
It’s unlikely that the market will open up any time soon so that you can get your LLM from someone who isn’t a fascist billionaire tech bro.
You can find open weight models (of all kinds, not just LLMs) freely available here: https://huggingface.co/
Many can be run on locally on gaming hardware. Some of the classification and object recognition models can be run on microcontrollers (a popular use it to use them to make aimbots in FPS games).
There are larger models that can be run on more expensive dedicated hardware (like NVIDIA DGX Spark, ~$5,000) if you want to keep your projects completely internal. Or you can run those open weight models on hardware hosted in datacenters ran by companies you research to be ethical.
You can use AI without using one of the AI companies.
Given equal scaling speed
This is already not the case for computers. And you’re trying to argue that we might one day be able to run the bigger models on consumer hardware, but I said it’s too expensive to train them. And you know that the cost isn’t just the RAM right?
Sure, my saying they will “always” be tied to concentrated capital is slightly hyperbolic. You never know. But it takes a whole lot of optimism to hope that LLMs will be decoupled from the billionaire class, hence why most analyses of the problems LLMs cause discuss them in this context.
This is already not the case for computers. And you’re trying to argue that we might one day be able to run the bigger models on consumer hardware, but I said it’s too expensive to train them. And you know that the cost isn’t just the RAM right?
If you replace everything I said about RAM with compute and replace the Bill Gates quote with Thomas Watson saying that the “world market [has room for] for maybe five computers” then it’s the same argument. They’re all ultimately built on technical advances in lithography and so growth in RAM and compute largely stem from similar sources.
Sure, my saying they will “always” be tied to concentrated capital is slightly hyperbolic. You never know. But it takes a whole lot of optimism to hope that LLMs will be decoupled from the billionaire class, hence why most analyses of the problems LLMs cause discuss them in this context.
I understand why it is done as shorthand.
I know this likely doesn’t apply to you, as you seem to be capable of understanding the nuance, it’s that a lot of people on social media (by people I’m including the bots) don’t make that distinction.
The same vitriol that is, rightfully, directed at OpenAI/Anthropic/Google/NVIDIA is also directed at things completely outside of the corporate LLM hellscape.
I’m mostly frustrated at seeing the sciences, researchers and open source AI projects catching strays and being painted with the same brush.
It’s useful shorthand among people who understand the nuance, but using that shorthand on social media invites misunderstanding and that is causing collateral damage.
Keep in mind that this only relates to the state of the largest LLMs. You can train generative models on consumer graphics cards. Object recognition, classification, etc are all tiny compared to LLMs. These are also AI […]
The linked Codeberg page refers to LLMs in their current, corporate-controlled shapIe. It is about using LLMs for coding.
Your argument is rather leading away from that.
Unfortunately “AI,” and LLMs in general are, at present, wholly inseparable from the people who use them, and especially, (and more pressingly), the
peoplecorporations that make them. This is the reason I maintain my opposition to “AI” (and LLMs in general) despite machine learning as a field and LLMs as technology being truly fascinating, and having immense potential.In my ideal world the corporations pushing “AI” wouldn’t exist, the datasets and recipes would be open, and model training would be a community led effort with Folding@Home style distributed compute used for training, pretty much removing the viability of “AI” companies. Without “AI” companies pushing for adoption, strong social pushback can be used to reign in the slopification of the commons. With those two together (and likely many other smaller changes elsewhere) LLMs can be seen for what they really are and their real pros and cons evaluated.
But that world isn’t currently feasible, and probably won’t ever be. Neither is toppling even a single corrupt corporation on the level of Alphabet, or Microsoft. So all we can do is evaluate whether we consider certain software’s use of “AI”/LLMs is acceptable, and advocate for not using them. Which is why I think efforts like openslopware, and pages like this are a positive thing, even if I don’t necessarily agree with some of the specifics.
You do realize LLMs are made outside the US as well. I have zero problem with Chinese companies making LLMs that are released in the open and you can even run on your own hardware.
Shitty corporations are a nigh universal constant, not something exclusive to America. That’s American exceptionalism talking.
The difference is that China actually deals with corporations when they act shitty as happened with Alibaba and Jack Ma because China has a dictatorship of the proletariat which creates different material conditions from the dictatorship of capital you have in burgerland. No exceptionalism needed to explain this, just minimal capacity for critical thought that some people lack.
as happened with Alibaba and Jack
I’m out of the loop, what happened to them?Never mind, googled it. I like how they regulate things.i wish someone could put the fear of god into my oligarchs. lol
It is lacking the “money problem”: the high cost is adsorbed by big capitalist organizations (at the cost of the hardware price inflating) that hope to get everyone addicted/dependent on the tool to then surge the prices.
This is not open tech, this is not for everyone, this is not to make a better world. This is business.
This is business.
So was the Opium trade.
The most recent release of SOTA models being both more intelligent and 20-30% cheaper is really fighting against this argument (Opus 5.5 & Sol 6).
20-30% cheaper is really fighting against this argument
Still not remotely enough to be profitable
Recent experience, developing in Opus 5.5. Spin up an agent to review the security aspects of the project, write a report. Pass the report to another Opus 5.5 agent and tell it to implement mitigations for confirmed problems found in the report. Implementation agent proceeds to make tests for the described exploits and gets shut down by the guardrails: Opus 5.5 is not permitted to “do that kind of work” had to scale back to Opus 4.8 for the implementation, though 5.5 was still happy to describe the chained attack scenarios, apparently it’s not allowed to develop demonstration code as it was trying to do for the test cases.
This is a pretty solid list, besides Code Quality. Most of the links there are user error. It’s akin to blaming a Junior dev for breaking production imo and undermines the entire page.
On one hand, many things that you think are “AI” are actually humans in another country pretending to be an AI chatbot
Since your target audience appears to be developers and software engineers, have there been any reported cases of human “AI” coding chat bots?
You know that would be a really great scam. Perfect way to get access to people’s computers
it’s really fascinating how one of the markers of such use is the speed at which new feature/capabilities are delivered.
in other words: if it seems too fast for a human to do, it’s automatically labeled as ai.
you know, license washing can work both ways
Wonder what kind if cost we’re looking at, when “washing” a library like chardet. And whether that’s something available to robin-hood stuff from the rich, or if it’s just the other way around.
there are already decomp efforts of proprietary games that are accelerated thanks to LLMs, so yes
Wow, I didn’t know. Are there any bigger projects out there doing this? Most I find when googling it, is a few YouTubers and people who do it small scale as some sort of tech demo.
I think it cuts both ways. A lead dev of the PS5 Linux effort recently quit over the number of AI-assisted “security researchers” trying to make a quick buck on bug bounties rather than contribute to open source efforts.
true, but those people imo are the equivalent of script kiddies(vibe kiddies?)
Slop kiddies, that’s what the developper called them
check out the mario kart wii decomp
Can the blind anti LLM jerk stop already it’s the same 2024 2025 arguments. Agentic coding has massively improved this year even local open source LLM’s are getting good
You talk about it being “blind” right under exhaustive list of reasons why.
The fact that these arguments appeared before doesn’t make them less true. Yes, the coding quality of LLMs has increased, and, notably, it stopped producing as many bugs all over the place, and even got to review bugs made by humans.
But it doesn’t remove all other issues from the table, namely:
- Issues of licensing and stolen code, potentially causing copyright problems which most open-source projects are poorly fit to resolve and may go down because of them
- Environmental impact of model training, huge data centers threatening water supplies, producing tons of e-waste, creating noise pollution, increasing power demand, including one from unclean sources at the critical point when we should reduce it as much as possible
- Deterioration in skill level of human coders and lack of avenues for juniors to learn to code properly, which is likely to stay as an essential thing to have
- Outside coding, there are many ethical issues around companies producing LLM models, as many have ties to the military and police, and are heavily misused to fuel fascist regimes
- And more.
Issues of licensing and stolen code, potentially causing copyright problems which most open-source projects are poorly fit to resolve and may go down because of them
What, we care about licenses and copyright now?
Big tech has stolen all code and now everything it “open source”. You’re not going to get into legal trouble for using LLM generated code there’s no precedent for that. The only ones getting sued are the big tech companies who stole the code.
Environmental impact of model training, huge data centers threatening water supplies, producing tons of e-waste, creating noise pollution, increasing power demand, including one from unclean sources at the critical point when we should reduce it as much as possible
Sure but that’s mostly them training a gazillion new models. You could ban models from providers that are doing all these things and thus promote more efficient models.
Deterioration in skill level of human coders and lack of avenues for juniors to learn to code properly, which is likely to stay as an essential thing to have
Valid but understanding architecture and delegating writing the code and especially tests is becoming much more important than just artisan writing it these days.
Outside coding, there are many ethical issues around companies producing LLM models, as many have ties to the military and police, and are heavily misused to fuel fascist regimes
Like I said there’s local models as well. China puts out plenty of great open weight stuff like GLM5.3 which will likely have to run in the cloud but it’s still open weight.
I see your point. I don’t think China-made local models are panacea by any means though, even through it’s good they try to make it more efficient instead of just throwing gazillions and watching the world burn. It is still a very massive, very problematic operation.
Running LLMs locally, especially considering the pace of development, is just going to distribute the e-waste issue and make parts availability even worse, as I doubt your average hardware would be busy around the clock as it is in data centers (and so we’ll need much more of it). It is better for some things (like privacy or reliability or screwing with incentives of those turning the economy and the world upside down for a quick grift), but it doesn’t solve the others.
As per legal aspects, time will be our judge. I’d love not to give a damn about copyright, but it still exists.
LLM’s can save tremendous time now. You see projects like C&C Generals running on a tablet, or Lego Racers on web which would require tonnes of manual labour and now they’re just a few LLM prompts. It’s gotten to the point these kind of projects which would have been multi-year undertakings in the past have become “unimpressive”
Yes they’re not perfect and they do make mistakes. That’s why understanding architecture is more important than ever. But manual coding is really becoming a thing of the past. Agentic coding progress over the last year has pretty much solidified that. It’s still not perfect but it’s really good now.
Any industry that throws billions of dollars into something will make progress and now that progress has been made there’s not much of a point in just resisting the tech itself. Therefore instead of pleading against all AI I believe it’s much better to plead for open weight or open source and efficient AI.
The problem is like with the Luddites that you don’t own the machine (AI) and the largest companies want to keep full control over them. It’s a problem with the system ruling you not with the technology itself.
The technology is undoubtedly good. All technologies are, if you ask me. The problem is about externalities, but they do matter and cannot be viewed separately.
Currently, as things stand, the development and, notably, use of the technology comes with a massive threat to several important areas. This is why many pieces of software that ban LLM contributions make a for now caveat.
Relying on it now, under current circumstances, is dangerous and damaging, even if sometimes it is a tremendous help. It may get better once we have truly efficient models and training, do not reinvent the wheel from scratch 20 times over, do not screw entire economy over it, put guardrails on the use of technology against people, and have a cheap, resource-efficient, and long-relevant hardware to run it. But we’re not there yet. We’re going to space with a gunpowder barrel here. Workable? I guess. Reasonable? No.
If the goal isn’t to oppose the technology but instead steer towards better use of the technology then that should be promoted. This blanket opposition means that we get fascists like DHH being the only ones to adopt LLM’s because for some reason the left has a hate boner for it.
We’ll stop when the AI evangelists finally learn the idea of consent and quit trying to force AI down everyone’s throats nonstop.
You want to shrivel up your brain with AI? Go ahead, you’re an adult (presumably) and I’m not your mom. Go access it through a browser or separate app. Quit trying to force it into my face by integrating it into the OS or DE. And while you’re at it, quit continually going “BuT dId YoU cOnSiDeR tHiS???1?” like a door-to-door salesman that jams their foot in the door so you can’t close it
This is the exact opposite. Trying to ban AI from everything despite its use cases.
Many people don’t know this but even Luddites didn’t oppose machinery. They simply opposed the ownership structure of it. So instead of saying “no LLM” you could say “Open-source/weight LLM’s allowed only”.
This is the exact opposite. Trying to ban AI from everything despite its use cases.
This is a pure, bold-faced lie. But this is nothing new from AI evangelists.
This is about banning AI contributions from specific open-sourced projects. This is due to a number of legitimate concerns that are still not solved, such as maintainability and how AI development legally interacts with licenses such as the GPL
You want to go and start an open sourced project that uses AI from the start? Go ahead, nothing is stopping you. You want to install AI programs on your OS? Go ahead, nobody is stopping you
But don’t cram that bullshit into the OS itself. Doing so then forces AI onto everybody whether they want it or not, and that’s bullshit
This is about banning AI contributions from specific open-sourced projects.
Do you even know what Codeberg did?
Stop trying to take performative outrage away from liberals, it’s literally their whole identity.
Do tankies tend to put “liberal” on anything they don’t like?
I’m sorry, comrade, but that’s not how it works. There are many valid reasons people of all political backgrounds stand against using LLMs in (open-source) coding.
Platforms and projects taking direct action on the matter is exactly what communists usually claim to support. They make change where they can, and rally for a bigger one.
Ultimately, the most performative folks are those judging others without making something better themselves.
Weird how the only reasons you see articulated are ones deeply rooted in liberalism though.
Ultimately, the most performative folks are those judging others without making something better themselves.
I guess we can agree on something. People screeching about AI without actually doing anything productive are the worst.
