Anarchism

if you have the means to operate them (tens to hundreds thousands of dollars worth of computer parts) plus maintenance you could probably do a lot even if you didn't run AI models on your fancy gear.

Just some things off the top of my head

Someone posted here recently on using tools to anonymize faces in the protest without being obvious about it, by face-swapping them. GenAI can do this easily.

For orgs with little resources, GenAI could help making more eye-catching flyers and posters but it needs to be counter-balanced by the risk of being labeled as "AI slop".

LLM tools can be used to detect and counter spam and shock campaigns in our social media.

Using it to disrupt fascists online is about the only thing I've come up with. They're doing it to us so 🤷‍♂️

Otherwise there's nothing AI can do that we can't already.

This is silly. Anarchist organizations are not different to any other organizations with respect to any data processing algorithms. AI is no different from other data processing algorithms with respect to any organizations save those selling AI.

Face recognition might be usable to spot agent provocateurs at protests but can be a double edged sword.

LLMs seem pretty useless to me and generative content AI seems to have very little utility as others here already said.

IMO, other than to fight against establishment AI-powered projects, avoid LLMs and image generation for public facing anarchist agitprop for now because it's currently associated with and controlled by the capitalist class. There isn't really a particularly great application area for LLMs and image generation where I would rather have a machine do the work than a human.

Optimal control theory and reinforcement learning can be used for peaceful robotics applications. Classical machine learning is used all the time in the background without controversy. We can keep doing that. IMO the CEOs and business people don't actually understand the mathematics behind the systems they work with, so if you like math and have the time to pore through research papers, you can expose the business people as jackasses who don't know shit about fuck (even more than they already do themselves).

In the background, work to wrest control of AI from the capitalist class if you have the resources and time. We should periodically revisit the issue as the culture changes and people learn how to use (and misuse) AI tools. IMO it makes sense to introduce it into anarchist spaces only when we can run our own AI resources independently from capitalists. But before that, we're also going to need some technically informed theory about the ethics and philosophy behind AI.

This might be an unpopular opinion, but I think we need more technical training on how to use AI, both the training/testing workflow and the mathematics behind it. In general, I think that we need to learn more about the technical tools of the oppressors, like missiles, UAVs, LRADs, surveillance technology, etc., so we can properly combat against them.