Simple: Because I am talking about LLMs.

You seem to think that AI is all statistical models. That is not true. Automated deduction, for example, contains no statistical reasoning at all. If it delivers results, these are reliable. AI is a very wide field. Many things get their own names when they start to work well, but they are still AI.

And what has that to do with my posting you answered to?

I guess they are just helping that along a bit...

That does not make any sense in the given context.

On a guess: This is about real (formal) deduction. Not the crap that some people mistake for deduction.

They have banned you from moderation?

Not quite. Misinformation requires active generation of content. You can get a similar temporary effect by letting deranged stupid people post and deleting anything that does not fit your desired misinformation pattern. But that is not sustainable since you will need to censor too much.

The math causes hallucinations. There is no way around that. It is statistics, not deduction. Deduction is immune to hallucination, but far out of computational reach. Statistics is within reach but has a fuzziness that cannot be removed.

Au contraire. People are fucking stupid. They just do not want to change anything because they fear understanding even less.

Lets hope the capacity to actually punish will not exist anymore at that time.

Other than that, why not start with a rapist, grifter, scammer and professional liar that thinks fascism is great and an ideal to aspire to? There is one active right at this moment....

They usually have the most "logical" arguments for their retardation and inaction. The rest of the world leaves these people behind eventually.

I think Schneier is far too generous here and still thinks these models can be really helpful for defenders. I do not think that is actually the case. They do help attackers up to intermediate levels, though.

But I might be mistaken. I currently have a student investigate some aspects of the question.

Moderation does generally not add things.

If your actually read my posting, you will see that it is actually "do not tust LLMs". And that is not a generalization at all, because for LLMs it is mathematically proven that hallucinations cannot be prevented.

Indeed. This fail on the part of Reddit is telling: They do not understand their business model! Nobody has any reason to go to Reddit except talking to actual humans.

My guess is they just "thought" that since everybody is doing "AI", they should so too.

That is for the crappy, unfixable LLM version, obviously. Other AI technologies can perform fine, but in entirely different tasks. The only somewhat reliable way the human race has for getting general answers is to ask an actual, human expert. Of course, fake "experts" are plentiful, and many people do not even have the basic fact-checking ability needed to separate real experts from fake ones. Explains a lot about the current AI hype and why democracy fails time and again.