Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

In Search of AI Psychosis

astralcodexten.com

Post reply on HN
Print view View original

131–140 of 258 posts

Re: In Search of AI Psychosis

#131

Post by reidalert »

The description of the risk factors very much jibes with what I have seen in a friend recently. He is quite isolated, and spends most of his evenings writing using AI (he works in a blue-collar trade and wouldn't be typing stuff out by hand usually).

He's convinced that he has discovered a grand theory of human connection / relationships / energy / physics, and keeps interrupting in conversation to explain how something I've said is just an example of a deeper pattern.

Sadly, this theory of connection is cutting him off from actual connection - he gets so much validation from AI that he believes he has discovered a new world model. But the people around him aren't bought into the vision (mostly because it is bullshit), and so he ends up even more isolated.

Re: In Search of AI Psychosis

#132

Post by gwd »

One problem with his survey -- The Lizardman Constant:

> Researchers have demonstrated repeatedly in human surveys the stylized fact that, far from being an oracle or gold standard, a certain small percentage of human responses will reliably be [nonsense]: “jokester” or “mischievous responders”, or more memorably, “lizardman constant” responders—respondents who give the wrong answer to simple questions.

> Below a certain percentage of responses, for sufficiently rare responses, much or all of responding humans may be lying, lazy, crazy, or maliciously responding and the responses are false. This systematic error seriously undermines attempts to study rare beliefs such as conspiracy theories, and puts bounds on how accurate any single survey can hope to be.

> For example, 4% of respondents may endorse the claim ‘lizard-people rule the earth’, 5% of atheists believe in God, a surprising number of adults believe you see by shooting beams from your eyes, and so on. This cautions us against taking survey results about extremely unusual people or traits too literally, or expecting perfectly accurate results, as given the lizardman constant and other crud factors, it is entirely possible that some or all of the outliers may just be the lizardman constant at work.

https://gwern.net/doc/sociology/survey/lizardman/abstract

Re: In Search of AI Psychosis

#133

Post by bawolff »

> Suppose that respondents had an average of fifty family members and co-workers, so that plus their 100 closest friends makes 150 people.

Say what now? Am i just really socially isolated? Seems insane to me to assume the average person is close enough to 150 people to know how much each of those 150 people use AI and if they are "psychotic".

Re: In Search of AI Psychosis

#136

Post by AstralStorm »

anon84873628 wrote: post #92
epiccoleman wrote: post #68

Earlier quoted context omitted.

> 'biological' mental illness is tightly coupled to qualitative mental state, and bidirectionally at that. Not only do your chemicals influence your thoughts, your thoughts influence your chemicals, and it's possible for a vulnerable person to be pushed over the edge by either kind of input. We like to think that 'as long as nothing is chemically wrong' we're a-ok, but the truth is that it's possible for simple norma…

>So it seems very spooky on its face that somehow thoughts themselves could have a bidirectional relationship with chemistry. There's no scientific reason to believe thoughts affect the chemistry at all. (Currently at least, but I'm not betting money we'll find one in the future). When Scott Alexander talks about feedback loops like bipolar disorder and sleep, he's talking about much higher level concepts. I don't re…

Sure the thoughts can influence your chemical state. Scott even provides an example. Suppose you become so engrossed in your weird idea you start to lose sleep over it... Or start to feel anxious about it.

At some point, your induced stress will cause relevant biological changes. Not necessarily directly.

PTSD indeed is likely an overload of a normal learning and stress mechanism.

Re: In Search of AI Psychosis

#137

Post by AstralStorm »

vintermann wrote: post #128
gherkinnn wrote: post #95

> First, much like LLMs, lots of people don’t really have world models. This is interesting and something I never considered in a broad sense. I have noticed how the majority of programmers I worked with do not have a mental model of the code or what it describes – it's basically vibing without an LLM, the result accidental. This is fine and perfectly workable. You need only a fraction of devs to purposefully shape t…

Everyone has a "world model". These models just differ on how much they care about various things. No one has a "world model" which literally encompasses everything about the world, that wouldn't be a model at all, it'd just be the world, much like a 1:1 map. Also, no one has a "world model" that is purely based on experiment and reason. Everyone gets their beliefs via other people first and foremost. Some get it fro…

For a lot of people, the world models are really rough and incomplete, so they really really on common opinion on these matters.

This is the same if you tried asking a general populace ethical questions in a vacuum sneakily. You're going to be dismayed after collecting the set of approved behaviors per culture.

There's not really a way to evaluate one of these.

Re: In Search of AI Psychosis

#139

Post by AstralStorm »

tuyiown wrote: post #118
gherkinnn wrote: post #95

> First, much like LLMs, lots of people don’t really have world models. This is interesting and something I never considered in a broad sense. I have noticed how the majority of programmers I worked with do not have a mental model of the code or what it describes – it's basically vibing without an LLM, the result accidental. This is fine and perfectly workable. You need only a fraction of devs to purposefully shape t…

You forgot the most important part, one's own model is not only probabilistic, it's also (more or less) forever challenged by reasoning to stabilize to some kind of self consistency. This refinement is critical and its mechanics still eludes everyone AFAIK.

Most people do not challenge theirs by reasoning, only by social approval - and that's easy to game.

That's why they turn 180 or radicalize badly when exposed to sufficiently strong social or usual media.

Re: In Search of AI Psychosis

#140

Post by AstralStorm »

andoando wrote: post #81
mk_stjames wrote: post #38

Earlier quoted context omitted.

It's actually not; there has been a phenomenon that Anthropic themselves observed with Claude in self-interaction studies that they coined 'The “Spiritual Bliss” Attractor State'. It is well covered in section 5 of [0]. >Section 5.5.2: The “Spiritual Bliss” Attractor State > The consistent gravitation toward consciousness exploration, existential questioning, and spiritual/mystical themes in extended interactions was…

I think a pretty simple explanation is that the deeper you go into any topic the closer you get to metaphysical questions. Ask why enough and you eventually you get to what is reality, how can we truly know anything, what are we, etc. It's a fact of life rather than anything particular and about llms

Normally people attempt to escape these time sinks when not explicitly taking about philosophy or they tend to elicit silence.

These are excellent nerd snipes though and for attmepting to make one sound profound to uneducated.

Post reply on HN