Hacker News Reader

Comment by l23k4 | Hacker News Reader

We had people acting out like this before LLM chatbots, correlation does not necessarily imply causation.

We did...but it was few here and there. The LLMs are making it massive and impacting people on a huge scale.

I'd like real numbers on that, because I don't see any evidence that this is particularly prevalent or novel issue. I see a lot of media hype on this topic, but that's not the same thing as actually being a real problem. Even the name is hype, ai psychosis is not a diagnosis a doctor would give here and none of these people really sound psychotic. They sound like people who believe in conspiracy theories and join cults.

> I see a lot of media hype

Right, a guy keeping endless stacks of paper with his ChatGPT-produced novelty "scientific" theories and wanting to "apply for pope" - definitely some "media hype" there.

People suffering from mental health issues were doing exactly that kind of things long before LLMs.

Delusions of grandeur are nothing new.

Who said they were new? But the ability to induce them ON SCALE is what is new, brmought to you by the LLM Slop companies.

Yeah, it's the "ON SCALE" claim I object to. I want evidence first; I'm not willing to accept the loss of freedoms that blindly accepting this premise would require.

We don't yet know if LLM chatbots are actually inducing them, or just offering a convenient outlet. Both options are perfectly feasible, although it seems unsurprising that people suffering from such delusions would seek out validation wherever they may find it.

Show me the evidence of the massive scale.

This is something new. Delusions were around before, certainly, but LLM offers a round the clock potential for psychological conditioning, which would not normally be possible without sustained attention by a group of people.

I don't know, internet already made it easy for delusional people to find each other (and trolls).

Look into the gangstalking/targeted individual communities, you can even frequently find these people on HN.

Flat earthers and sovereign citizens are two other famous examples. People also seem to regularly move between these groups, the specific delusional belief probably isn't the point.

I had an encounter with someone who thought he was being gangstalked (by me and others). He made rude and threatening comments to me and others. He accused me of being a double on one occasion and a police man on another. When I complained about what he said, he asked me why I was being defensive. However, I would argue that he suffered from a pre-existing psychological condition probably accelerated by being homeless and being badly treated in the past. It became a self-fulfilling prophecy.

As for Flat Earth theory, it has been heavily promoted by mainstream media in the last decade or two after years of obscurity. It is a useful strawman for the ruling class to associate dissent with. Along with the tinfoil hat meme (I have never met anyone who wears one. It's just another image of the extreme to promote group conformity.)

> correlation does not necessarily imply causation

I feel like you're missing what you're replying to, why are you saying this? The article is about a person who "lost grip on reality", no one is saying LLMs is turning people into pope-wannabees as far as I can tell, you're reacting against something no one claimed.

Explicit accusation that this was caused by chatbots + call for general regulation is right there in the article:

"AFP spoke to several members about their experiences. All warned that the world has to wake up to the threat unregulated AI chatbots pose to mental health.

Questions are also being asked about whether AI companies are doing enough to protect vulnerable people."

This, in time, might be used to nerf the models that we use. Of course, one actor is singled out:

"There has also been a recent rise in people spiralling while using Elon Musk's xAI's Grok chatbot, he said."

I don't think "correlation does not necessarily imply causation" even makes sense to someone saying "Maybe AI chatbots aren't great for people's mental health" or even "Are the AI companies actually trying to prevent AI chatbots being bad for people's mental health?", both statements seem fine and doesn't imply any causation as far as I understand.

This cautious statement, which would indeed be fine, is an invention of yours when it comes to the article. They assert causation, calling for AI companies to be disciplined and punished, praising the EU online censorship campaign, arguing this is a big experiment:

"Millar called for AI companies to be held responsible for the impact of their chatbots, saying the European Union has been more assertive in regulating Big Tech than the US or Canada.

He believes spirallers like him have unwittingly been caught in a massive global experiment."

They're calling for "AI companies to be held responsible for the impact of their chatbots" which regardless of what happened before, sounds like a reasonable thing to do, you don't even need to try to look at any correlation or causation to arrive at this.

I still don't see where the whole "correlation does not necessarily imply causation" comes in, so because this person was personally affected, they shouldn't reach the conclusion that AI companies need to be held responsible for whatever effects they have?

"you don't even need to try to look at any correlation or causation to arrive at this"

You actually very much do. If there is no causation, then it is not "their impact". Otherwise, you're just advocating for rope-makers to be punished for wrong-use of their product. Which is, of course, foolish.

Ok, put another way, do you think we clearly know exactly the pros and cons for individuals to use AI chatbots for mental health treatment?

If you don't clearly know exactly those things, wanting anything else than "AI companies to be held responsible for the impact of their chatbots" as a first step would be utterly foolish and de-humanizing.

> "AI companies to be held responsible for the impact of their chatbots"

Sure, but people on HN have a weird idea of what "held responsible" means. Often it seems to mean "harshly punished", as opposed to making the injured parties whole.

This probably reflects a split between US culture and the rest of the Western world.

Using uncertainty to punish your perceived enemies is called "injustice".

No, figuring out harms is called "thinking before acting", and unless you want laws and regulations written in blood (something I personally want to avoid), you need to think before acting.

Figuring out what harms something in wide use has, is a good thing, and doesn't mean "ban it today", it means "lets figure it out".

Language this vague has a name: it's called "bullshit".

Reminds me of another slang, probably applicable to this entire discussion: "pointless"

Reply on news.ycombinator.com