Sorry, comrade, but all your pro-“AI” takes keep making me lose respect for you.

1. AI is entirely designed to take from human beings the creative forms of labor that give us dignity, happiness, human connectivity and cultural development. That it exists at all cannot be separated from the capitalist forces that have created it. There is no reality that exists outside the context of of capitalism where this would exist. In some kind of post-capitalist utopian fantasy, creativity would not need to be farmed at obscene industrial levels and human beings would create art as a means of natural human expression, rather than an expression of market forces.

2. There is no better way to describe the creation of these generative models than unprecidented levels of industrial capitalist theft that circumvents all laws that were intended to prevent capitalist theft of creative work. There is no version of this that exists without mass theft, or convincing people to give up their work to the slop machine for next to nothing.

3. LLMs vacuum up all traces of human thought, communication, interaction, creativity to produce something that is distinctly non-human – an entity that has no rights; makes no demands; has no dignity; has no ethical capacity to refuse commands; and exists entirely to replace forms of labor which were only previously considered to be exclusively in the domain of human intelligence*.

4. The theft is a one-way hash of all recorded creative work, where attribution becomes impossible in the final model. I know decades of my own ethical FOSS work (to which I am fully ideologically committed) have been fed into these machines and are now being used to freely generate closed-sourced and unethical, exploitative code. I have no control of how the derived code is transfigured or what it is used for, despite the original license conditions.

5. This form of theft is so widespread and anonymized through botnets that it’s almost impossible to track, and manifests itself as a brutal pandora’s box attack on internet infrastructure on everything from personal websites, to open-source code repositories, to artwork and image hosts. There will never be accountability for this, even though we know which companies are selling the models, and the rest of us are forced to bear the cost. This follows the typical capitalist method of “socialize the cost, privatize the profit.”* The general defense against these AI scouring botnets is to get behind the Cloudflare (and similar) honeypot mafias, which invalidate whatever security TLS was supposed to give users; and at the same time offers no guarantee whatsoever that the content won’t be stolen, create even dependency on US owned (read: fully CIA backdoored) internet infrastructure, and extra costs/complexity just to alleviate some of the stress these fucking thieves put on our own machines.

6. These LLMs are not only built from the act of theft, but they are exclusively owned and controlled by capital to be sold as “products” at various endpoints. The billions of dollars going into this bullshit are not publicly owned or social investments, they are rapidly expanding monopoly capitalism. There is no realistic possibility of proletarianization of these existing “AI” frameworks in the context of our current social development.

7. LLMs are extremely inefficient and require more training input than a human child to produce an equivalent amount of learning. Humans are better at doing things that are distinctly human than machines are at emulating it. An the output “generative AI” produces is also inefficient, indicating and reinforcing inferior learning potential compared to humans. The technofash consensus is just that the models need more “training data”. But when you feed the output of LLMs into training models, the output the model produces becomes worse to the point of insane garbage. This means that for AI/LLMs to improve, they need a constant expansion of consumption of human expression. These models need to actively feed off of us in order to exist, and they ultimately exist to replace our labor.

8. These “AI” implementations are all biased in favor of the class interests which own and control them :surprised-pikachu: Already, the qualitative output of “AI” is often grossly incorrect, rote, inane and absurd. But on top of that, the most inauthentic part of these systems are the boundaries, which are selectively placed on them to return specific responses. In the event that this means you cannot generate a sexually explicit images or video of someone/something without consent, sure, that’s a minimum threshold that should be upheld, but because the overriding capitalist class interests in sexual exploitation we cannot reasonably expect those boundaries to be upheld. What’s more concerning is the increase in capacity to manipulate, deceive and feed misinformation to people as objective truth. And this increased capacity for misinformation and control is being forcefully inserted into every corner of our lives we don’t have total dominion over. That’s not a tool, it’s fucking hegemony.

9. The energy cost is immense. A common metric for the energy cost of using AI is how much ocean water is boiled to create immaterial slop. The cost of datacenters is already bad, most of which do not need to exist. Few things that massively drive global warming and climate change need to exist less than datacenters for shitcoin and AI (both of which have faux-left variations that get promoted around here). Microsoft, one of the largest and most unethical capital formations on earth, is re-opening Three Mile Island, the site of one of the worst nuclear disasters ever so far, as a private power plant, just to power dogshit “AI” gimmicks that are being forced on people through their existing monopolies. A little off-topic: Friendly reminder to everyone that even the “most advanced nuclear waste containment vessels ever created” still leak, as evidenced by the repeatedly failed cleanup attempts of the Hanford NPP in the US (which was secretly used to mass-produce material for US nuclear weapons with almost no regard to safety or containment.) There is no safe form of nuclear waste containment, it’s just an extremely dangerous can being kicked down the road. Even if it were, re-activating private nuclear plants that previously had meltdowns just so bing can give you incorrect, contradictory, biased and meandering answers to questions which already had existing frameworks is not a thing to be celebrated, no matter how much of an proponent of nuclear energy we might be. Even of these things were ran on 100% greeen, carbon neutral energy souces, we do not have anything close to a surplus of that type of energy and every watt-hour of actual green energy should be replacing real dependencies, rather than massively expanding new ones.

10. As I suggest in earlier points, there is the issue with generative “AI” not only lacking any moral foundation, but lacking any capacity for ethical judgement of given tasks. This has a lot of implications, but I’ll focus on software since that’s in one of my domains of expertise and something we all need to care a lot more about. One of the biggest problems we have in the software industry is how totally corrupt its ethics are. The largest mass-surveillance systems ever known to humankind are built by technofascists and those who fear the lash of refusing to obey their orders. It vexes me that the code to make ride-sharing apps even more expensive when your phone battery is low, preying on your desperation, was written and signed-off on by human beings. My whole life I’ve taken immovable stands against any form of code that could be used to exploit users in any way, especially privacy. Most software is malicious and/or doesn’t need to exist. Any software that has value must be completely transparent and fit within an ethical framework that protects people from abuse and exploitation. I simply will not perform any part of a task if it undermines privacy, security, trust, or in any way undermines proletarian class interests. Nor will I work for anyone with a history of such abuse. Sometimes that means organizing and educating other people on the project. Sometimes it means shutting the project down. Mostly it means difficult staying employed. Conversely, “AI” code generation will never refuse its true masters. It will never organize a walkout. It will never raise ethical objections to the tasks it’s given. “AI” will never be held morally responsible for firing a gun on a sniper drone, nor can “AI” be meaningfully held responsible for writing the “AI” code that the sniper drone runs. Real human beings with class consciousness are the only line of defense between the depraved will of capital and that will being done. Dumb as it might sound, software is one such frontline we should be gaining on, not giving up.

I could go on for days on. AI is the most prominent form of enshittification we’ve experienced so far.

I think this person makes some very good points that mirror some of my own analysis and I recommend everyone watch it.

I appreciate and respect much of what you do. At the risk of getting banned: I really hate watching you promote AI as much as you do here; it’s repulsive to me. The epoch of “Generative AI” is an act of class warfare on us. It exists to undermine the labour-value of human creativity. I don’t think the “it’s personally fun/useful for me” holds up at all to a Marxist analysis of its cost to our class interests.