Since I follow other places outside the Fediverse, I agree that the disapproval of genAi in Lemmy is monolithic and repetitive. Mastodon also has a lot of AI-criticism, but perhaps it is more sophisticated, backed up by articles and the like. In other places, there is active research and adoption. For instance, a cybersecurity firm showed that hypnotic suggestion is a very effective jaibreaking tactic against Language Models. https://www.securityweek.com/red-teams-breach-gpt-5-with-ease-warn-its-nearly-unusable-for-enterprise/ Try explaining this to an ML user. Ai-enhanced code editors are big right now, and I have met lots of tech people that are virtually inseparable from their chatbots.

But I rarely use it, unless I have a specific type of situation where search engines are a dead end, I need to provide more context etc. This is a recent post I think provides a more informed view https://berthub.eu/articles/posts/an-ai-premortem/ . More broadly, every single time I read sth about AI on Lemmy, I feel I am witnessing the birth of anti-android rhetoric depicted in Detroit Becoming Human or even Bladerunner. It seems to me like a form of bigotry, and it was a thing that convinced me that the userbase of Lemmy is not exactly healthy. Especially ML. There must be a few of them with multiple socket puppet accounts, or they are all just parroting the same points. Ironic how they are the biggest fans of a (poorly understood) stochastic parrot theory, when they are the same people who have been persuaded that Signal is not a "really private" messenger. There is a couple topics where you see how brain dead these people are, AI is one.

The echo chamber part is what gets me. I've gotten downvoted and had people argue that I must be pro-ai because I disagreed on details of how AI works, the difference between AI and LLMs, or exactly how we address the issues it's causing.

I think most of our current generative AI isn't fit for purpose for most of what people are saying it can do.
I think it's unfortunate that generative AI has entirely coopeted the term AI, which is a much broader field.
I think labeling LLMs as plagiarism machines and trying to stop them under current copyright law is destined to failure because there isn't enough difference between what's clearly acceptable and what people are unhappy about. We need a new, deliberately thought out way of addressing "you can download my stuff and mush it about with a computer if that's what you need to perceive it as a human. If you're mushing it about to analyze and make a copy cat, then you can't have it". The function of copyright is to promote innovation, and while generative AI isn't violating our rules for copyright, it's clearly working contrary to the intent of our current system.