Technology

I think OpenAI’s recent sycophant issue has cause a new spike in these stories. One thing I noticed was these observations from these models running on my PC saying it’s rare for a person to think and do things that I do.

The problem is that this is a model running on my GPU. It has never talked to another person. I hate insincere compliments let alone overt flattery, so I was annoyed, but it did make me think that this kind of talk would be crack for a conspiracy nut or mentally unwell people. It’s a whole risk area I hadn’t been aware of.

https://www.msn.com/en-us/news/technology/openai-says-its-identified-why-chatgpt-became-a-groveling-sycophant/ar-AA1E4LaV

From the article (emphasis mine):

Having read his chat logs, she only found that the AI was “talking to him as if he is the next messiah.” The replies to her story were full of similar anecdotes about loved ones suddenly falling down rabbit holes of spiritual mania, supernatural delusion, and arcane prophecy — all of it fueled by AI. Some came to believe they had been chosen for a sacred mission of revelation, others that they had conjured true sentience from the software.

/.../

“It would tell him everything he said was beautiful, cosmic, groundbreaking,” she says.

From elsewhere:

Sycophancy in GPT-4o: What happened and what we’re doing about it

We have rolled back last week’s GPT‑4o update in ChatGPT so people are now using an earlier version with more balanced behavior. The update we removed was overly flattering or agreeable—often described as sycophantic.

I don't know what large language model these people used, but evidence of some language models exhibiting response patterns that people interpret as sycophantic (praising or encouraging the user needlessly) is not new. Neither is hallucinatory behaviour.

Apparently, people who are susceptible and close to falling over the edge, may end up pushing themselves over the edge with AI assistance.

What I suspect: someone has trained their LLM on somethig like religious literature, fiction about religious experiences, or descriptions of religious experiences. If the AI is suitably prompted, it can re-enact such scenarios in text, while adapting the experience to the user at least somewhat. To a person susceptible to religious illusions (and let's not deny it, people are suscpecptible to finding deep meaning and purpose with shallow evidence), apparently an LLM can play the role of an indoctrinating co-believer, indoctrinating prophet or supportive follower.

This happened to a close friend of mine. He was already on the edge, with some weird opinions and beliefs… but he was talking with real people who could push back.

When he switched to spending basically every waking moment with an AI that could reinforce and iterate on his bizarre beliefs 24/7, he went completely off the deep end, fast and hard. We even had him briefly hospitalized and they shrugged, basically saying “nothing chemically wrong here, dude’s just weird.”

He and his chatbot are building a whole parallel universe, and we can’t get reality inside it.

I think that people give shows like the walking dead too much shit for having dumb characters when people in real life are far stupider

I lost a parent to a spiritual fantasy. She decided my sister wasn't her child anymore because the christian sky fairy says queer people are evil.

At least ChatGPT actually exists.

In that sense, Westgate explains, the bot dialogues are not unlike talk therapy, “which we know to be quite effective at helping people reframe their stories.” Critically, though, AI, “unlike a therapist, does not have the person’s best interests in mind, or a moral grounding or compass in what a ‘good story’ looks like,” she says. “A good therapist would not encourage a client to make sense of difficulties in their life by encouraging them to believe they have supernatural powers. Instead, they try to steer clients away from unhealthy narratives, and toward healthier ones. ChatGPT has no such constraints or concerns.”

This is a rather terrifying take. Particularly when combined with the earlier passage about the man who claimed that “AI helped him recover a repressed memory of a babysitter trying to drown him as a toddler.” Therapists have to be very careful because human memory is very plastic. It's very easy to alter a memory, in fact, every time you remember something, you alter it just a little bit. Under questioning by an authority figure, such as a therapist or a policeman if you were a witness to a crime, these alterations can be dramatic. This was a really big problem in the '80s and '90s.

Kaitlin Luna: Can you take us back to the early 1990s and you talk about the memory wars, so what was that time like and what was happening?

Elizabeth Loftus: Oh gee, well in the 1990s and even in maybe the late 80s we began to see an altogether more extreme kind of memory problem. Some patients were going into therapy maybe they had anxiety, or maybe they had an eating disorder, maybe they were depressed, and they would end up with a therapist who said something like well many people I've seen with your symptoms were sexually abused as a child. And they would begin these activities that would lead these patients to start to think they remembered years of brutalization that they had allegedly banished into the unconscious until this therapy made them aware of it. And in many instances these people sued their parents or got their former neighbors or doctors or teachers whatever prosecuted based on these claims of repressed memory. So the wars were really about whether people can take years of brutalization, banish it into the unconscious, be completely unaware that these things happen and then reliably recover all this information later, and that was what was so controversial and disputed.

Kaitlin Luna: And your work essentially refuted that, that it's not necessarily possible or maybe brought up to light that this isn't so.

Elizabeth Loftus: My work actually provided an alternative explanation. Where could these merit reports be coming from if this didn't happen? So my work showed that you could plant very rich, detailed false memories in the minds of people. It didn't mean that repressed memories did not exist, and repressed memories could still exist and false memories could still exist. But there really wasn't any strong credible scientific support for this idea of massive repression, and yet so many families were destroyed by this, what I would say unsupported, claim.

The idea that ChatBots are not only capable of this, but that they are currently manipulating people into believing they have recovered repressed memories of brutalization is actually at least as terrifying to me as it convincing people that they are holy prophets.

Edited for clarity

I read the article. This is exactly what happened when my best friend got schizophrenia. I think the people affected by this were probably already prone to psychosis/on the verge of becoming schizophrenic, and that ChatGPT is merely the mechanism by which their psychosis manifested. If AI didn’t exist, it would've probably been Astrology or Conspiracy Theories or QAnon or whatever that ended up triggering this within people who were already prone to psychosis. But the problem with ChatGPT in particular is that is validates the psychosis… that is very bad.

ChatGPT actively screwing with mentally ill people is a huge problem you can’t just blame on stupidity like some people in these comments are. This is exploitation of a vulnerable group of people whose brains lack the mechanisms to defend against this stuff. They can’t help it. That’s what psychosis is. This is awful.

... then they are not losing much

I admit I only read a third of the article.
But IMO nothing in that is special to AI, in my life I've met many people with similar symptoms, thinking they are Jesus, or thinking computers work by some mysterious power they posses, but was stolen from them by the CIA. And when they die all computers will stop working! Reading the conversation the wife had with him, it sounds EXACTLY like these types of people!
Even the part about finding "the truth" I've heard before, they don't know what it is the truth of, but they'll know when they find it?
I'm not a psychiatrist, but from what I gather it's probably Schizophrenia of some form.

My guess is this person had a distorted view of reality he couldn't make sense of. He then tried to get help from the AI, and he built a world view completely removed from reality with it.

But most likely he would have done that anyway, it would just have been other things he would interpret in extreme ways. Like news, or conversations, or merely his own thoughts.

Oh wow. In the old times, self-proclaimed messiahs used to do that without assistance from a chatbot. But why would you think the "truth" and path to enlightenment is hidden within a service of a big tech company?

TLDR: Artificial Intelligence enhances natural stupidity.