I expect that AI will increase the need for human moderation. LLMs are better at producing passable spam than identifying it, so any gains in AI moderation will be more than offset. I actually worry that AI spam will become too powerful for even good human+AI moderation and we'll see the death of high quality open discussions online.