Because the base arguments against AI are:
It produces slop. Sometimes entertaining slop, sometimes marginally useful slop, but slop.
It cannot produce anything better than slop. This is, perhaps, the key point. The proponents claim otherwise, without any evidence to support them.
Some of the slop is actively dangerous. (See: Suggestions of safe mushrooms, inaccurate legal advice, and whatever the hell is going on with the character chatbots that teens are talking with.)
People who auto-produce slop, fail to learn the skills they would have developed by writing/drawing/coding/singing/etc.
Gen-AI is utterly useless as a teaching tool. It is worse than useless: it actively discourages people from developing their own skills.
It's an incredible resource drain. Exactly how bad gets debated, but the point is: In order to have AI slop not be the random 5-legged horse with green mush tree backgrounds, you need a LOT of computer power.
The only reason we have the current level of entertaining-to-useful slop (it's still slop) is venture capitalists sinking billions into a pipe dream that is never going to happen. AI will never produce insightful, entertaining fiction; it will never produce a coherent movie; it will never produce blueprints for a house or an Ugly Christmas Sweater knitting pattern. The "hallucinations" are never going away; they're part of the process.
Again: Proponents claim otherwise, despite computer experts saying, no, that isn't how this can work.
Part of that "investment" includes widescale data & content scraping, some of which is definitely illegal, much of which is immoral and might be illegal, and, importantly, they are very aware that paying market value for the content they're using would tank any chance of profitability.
There might be useful potential aspects to gen-AI. (The same way there might be potentially good uses for blockchain.) Those are not the aspects the venture capitalists are sinking money into, because those are not "replace human creativity with a robot" features.
And once the pro-AI contingent has loudly declared that none of that matters and the haters are wrong because any day now, AI is going to be able to make apps that meet HIPAA security standards, or good knitting patterns based on particular types of yarn, or a chatbot that can act as a therapist without suggesting you get divorced or commit suicide, or beautiful artwork that doesn't hit the uncanny valley and definitely isn't copyright infringement, I promise bro, just wait, the NEXT version will fix this problem...
The anti-AI contingent has to fall back on "AI is stupid and it sucks and I hate it" because the people who are invested in it succeeding have blithely ignored, not refuted, every single problem with it.
The most reasonable analyses I've seen, say we've got another 2-3 years of it at most before the bubble collapses.
Because, and again, this is important:
It cannot produce better than slop.
There can be value in slop. The billions being thrown into it are not based on "let's find the best use for the kinds of slop this can make."
