[impossiblefork]

Personally the way I'm thinking is AI cooperatives. Democratic membership controlled organisations with relatively low fees to train AI aligned with the interests of the members.

Such organisations won't solve the fact that functioning AGI or very effective LLMs that can program or solve general text tasks precisely reduces the power of workers, but they'd at least mean that ordinary people control their own propaganda apparatus and are not stuck with models that are part of other people's propaganda apparatus (the state, Microsoft shareholders, some charity board that they do not control etc.).