AI Psychosis - Please take this seriously

In My Humble Opinion
41

So to be clear, you’re now accusing me of lying, right? You think that I’m not writing my posts. You’d rather accuse me of being a liar, despite having a trivially easy time of just… looking for yourself. Search for my name, search for Great Debates, search for the years 2005-2020. You will see my posting style and length has been consistent this whole time. But, of course, why put in a trivial amount of effort when you can instead imply someone is a liar?

You are also accusing me of not only being a liar, but engaging in premeditated faking the provenance of my original post in this thread, given that I actually posted a transcript that showed it was a fully formed argument before I asked for any feedback from an AI. Because how could my story not add up unless I specifically had another AI write it (or claude in another window), and then fake starting a conversation with Claude with my fully formed argument, unless I was deliberately planning to make a deception from the start?

42

I don’t think @Maserschmidt is accusing you of lying, or of posting AI generated text. However, they asked “How much [of these very long posts] are being shaped by chatbots?” And you admitted that sometimes your posts are indeed “shaped by chatbots,” at least by my interpretation of that phrase.

You said “I ask them for feedback and discuss the topics with them. Sometimes they’ll proofread.” That sounds like shaping to me. Not the text itself, but the content. I fully believe you structure your own posts and are writing the words themselves.

I’ll be honest, and I know you’re not going to care but I’ll say it anyway – I don’t have much interest in conversing with someone who’s bouncing all of their ideas of a chatbot first. If you’d like me to explain why I can; but know that I’ve talked to other humans and a lot of us feel the same way. I guess what you do with this opinion is up to you, but to me, it lessens your impact as a poster. 2 cents for your pocket.

43

I think you’ve got something closer to “AI psychosis” than I do. Now anyone who discusses any ideas with a chat bot is not worth talking to? If I think deeply about a topic, listen to feedback, ask where my ideas are weak, where they can be improved, and come here with a better argument, somehow what I have to write is contaminated and unworthy of human consumption? That’s, at the very least, a clear use of the genetic fallacy.

I have actually been incredibly transparent in this regard because I posted the actual conversation I had. And @Maserschmidt, who accuses me of using me of letting AI write my posts because they’re long, is not going to put in any effort to follow up on my request to look at what I was writing on this board 15 years ago either and see if it’s any different.

Easier to not actually read the available information that I’ve offered, and accuse me of having “AI psychosis” or lying about using AI to write my posts. Why engage with my words and my ideas when you can simply entirely dismiss someone who challenges you by declaring them to be contaminated and unworthy of discussion?

I use LLMs. They’re an interesting tool. They often teach you things and challenge your ideas. They’re highly modified by incentives. There’s a garbage in/garbage out principle and you get really good results from them by being thoughtful about engineering how you speak to them. I do this with my human friends too all the time because, as you might guess from my 25 year presence on these message boards, I like to discuss things and have my views challenged.

You guys are engaging in groupthink - AI is bad, anyone who talks to AI is unworthy of being interacted with and probably has “AI psychosis” and is hopeless contaminated. That’s a moral panic at best. You are not upholding the highest principles of debate here but I suspect you think you are. There’s a righteousness that usually comes with moral panics.

It’s funny that you started this thread with “please take this seriously” but you’re not taking me seriously. The funny thing about the knee jerk anti-AI crowd is that 90% of people have the exact same view as you, but you’ve all somehow convinced yourselves that you’re the unique rebels fighting against AI and everyone else has fallen for it. It’s one of those peculiar paradoxical situations where the vast majority of people have somehow convinced themselves that their overwhelmingly common view is unique and rebellious. You’re not being brave by making a post about “AI psychosis” – if anything I’m being brave by being willing to say anything other than the AI is the worst thing ever and ruins anything it touches. Because I guarantee you that I get a thousand times more social punishment for being curious about AI than you get for being concerned about it and this thread is proof.

For what it’s worth, I didn’t consult any AI in any post I wrote since the first one I wrote in this thread, but I guess I’m so hopelessly contaminated that it doesn’t matter. I cannot possibly be thinking for myself anymore.

44

So I said, “If you’d like me to explain why, I can.” You expressed zero curiosity in your reply, no attempt to understand my reasoning.

Why?

45

So – you tell me you’re not interested in talking to me, that I’m hopelessly contaminated and unworthy of talking to, and then you expect me to ask you why you’ve made this decision when you’ve already rendered your negative judgment towards me and said you didn’t want to talk to me?

I offered you information proactively. I’m not going to play your little game of trying to pull your explanation out of you, which you deliberately chose not to offer, after you’ve already indicated you have no interest in talking to me. That’s an absurd social dynamic that you tried to set up.

Why?

I did try to understand your reasoning. You chose your words and what to offer and I responded to it. Saying “no, you should’ve asked me what I really thought!” is a bizarre game that I’m not going to play with you.

46

Sincerely, you’re putting words in my mouth. I said, “I don’t have much interest in conversing with someone who’s bouncing all of their ideas off a chatbot first.”

I did not say “I’m not interesting in talking to you.” I did not say that you’re hopelessly contaminated and unworthy of talking to.

I’ve provided cites, you’ve provided none. You’ve misrepresented what Maserschmidt has said, and now you’re putting words in my mouth.

Am I misreading the situation?

47

Yes, you are clearly misreading the situation. Have you ever done peer reviewed research? Do you understand how to read it? Did you read the citations you provided, or just the abstract? Do you know how they chose to operationalize their definitions? Do you know what the research question is and their strategy for asking it? Do you know the ecological validity of what they’re aiming to explain? Do you know how to validate their methodology or their statistics? Do you know what the limits of what they’re attempting to claim are? Or did you just try to find an abstract that seems to say what you wanted to say and drop it into the conversation as though it’s an argument winning weapon?

I’m actually engaging with what you have to say. I’m also sharing examples of what I’m trying to demonstrate, which you refused to read. I challenged you to do it yourself to compare for yourself, which you refused to do. You think a 30 second google search and finding an abstract that seems to agree with you that you can drop in as an abstract is somehow superior to actually reading the evidence I’m presenting or simply gathering your own. Ah, yes, but I’m not engaging in peer reviewed research, so obviously you win by trumping me with a link to a paper you haven’t read. All I’m doing is actually engaging with the concepts we’re talking about and trying to demonstrate them to you and challenging you to test them yourself. There’s no way that’s as worthy as 30 seconds on google scholar citing a study you didn’t read and don’t understand.

You’re not considering what I have to say. You’re unwilling to read the evidence I provide or accept my challenge to look yourself. If I dropped some sort of citation, you wouldn’t read that either. Your conclusion is already made.

48

Wow.

I spent way more than 30 seconds finding those cites. I read through the actual papers where I could (there was only one where I could only get the abstract, and I highlighted that out of academic honesty. Scroll up, you’ll see it.) I read the rest to find out their methodology, how they defined sycophancy, and which models they used.

You are simply wrong about me and how I’m approaching this conversation. If you have a problem with the cites, with their methodology, then feel free to attack them. Find rebuttal cites. I WILL READ THEM! I also have a long history on this board, and you will find that I love reading cites.

You shared ONE example, an anecdote. I spent about 5 minutes reading through it, but it was LONG. And repetitive. And most of it wasn’t your words but a chatbot’s. I read every single word that YOU WROTE in that conversation, I just skimmed most of the replies by Claude.

If you don’t believe me, fine, but this is how I always operate. You’re simply wrong.

49

And you’re not interested in investigating it yourself? To see if there’s a meaningful difference between models that were thoughtfully designed for anti-sycophancy and models that are enthusiastically sycophantic? Ah - I guess you can’t. Because then you’d be contaminated, and there’d be no reason for anyone to talk to you.

50

What do you mean? I investigated! That’s why I found the cites! I used Chatbots extensively for years. I was a software engineer up until 6 months ago, do you think that I wasn’t using AI daily?

I also read the link you posted! I gave it a fair shot! You said you understood that people were reluctant to read that much text, especially generated text. I agree! I’m reluctant, but I made a good faith effort.

51

When you use gendered pronouns to refer to a LLM you have a problem.

53

Moderating:

Everyone calm down please. Too many exchanges are too close too attacking the poster and not the post.

It looks like you had a good discussion going that is now getting very heated. Please dial it down a bit.

Moderators and assumptions of good/bad faith
54

This topic was automatically opened after 60 minutes.

55

Oh please. Like the millions of people who call Siri or Alexa “she”? When you give something a gendered name, people tend to thoughtlessly default to using a gendered pronoun. I’m not confused about whether LLMs are people, and jumping to pathologize me (you didn’t even say “you may have a problem”) is silly.

Anyway, I think you guys have formed a pretty self sealing view at this point that’s unfalsifiable. If I argue, it’s more proof that I have AI psychosis. So I’m dipping out. Ping me if there’s something you think I should respond to.

Moderators and assumptions of good/bad faith
56

I didn’t mention you.

57
hidden as started with a dig ignoring the recent modnote

The main thing is that you started taking things personally and getting offended and even hateful. At that point, your posts became unconvincing.

I have had things I say to people shaped by AI, but not in general. I use it when there is conflict and I had to take a step back, and want to try to think through how to resolve things. When things are touchy, the exact wording can matter, and I take advantage of its “sycophancy” to help me word things in ways that are less hurtful, asking it how I might be misunderstood, and stuff like that.

So do I think doing that indicates any sort of psychosis? Of course not. I do notice that it can become easy to get caught up in, but so can anything. That’s not what AI psychosis is.

It specifically refers to mental illness that is triggered or exasperated by usage of these LLMs. Sure, everyone it happens with will have some sort of preconditions. It’s not inherent in the system.

Looking it up, it does seem that Anthropic did start trying to build things into Claude to detect issues. Initially, at least, it was quite crude and pissed people off. Hopefully they will get better.

I agree with you that the solution is gonna be safeguards and education. I do think maybe some degree of regulation may be relevant in that, but banning is not the answer.

The emotions AI stirs up bother me in both directions. Both the people who treat it like the devil, but also those who seem to get attached to it in ways that seem somewhat parasocial. People who make it part of their identity.

Neither of these is psychosis. It’s just the same thing that happens with video games, movies, and such. Maybe AI is more addictive in some ways, but then there are people who get addicted to video games.

Psychosis is something more. Something closer to what the OP described in their friend. Where they seem to be isolating and becoming intolerant of AI being wrong. Where they are starting to believe the AI over reality.

And even that seems more “danger signs” than fully psychosis yet.

disclosure: i have never made a post on this board that was shaped by AI. I have looked up something on Google and used the AI response, but only on things I already know about and just needed a refresher–like not remembering a word, or something. If I don’t know things, I look at the sources.

Moderators and assumptions of good/bad faith
Moderators and assumptions of good/bad faith
59

Moderating
This is really pushing into attacking the poster.
I probably should bump this up to a warning.

Back off and get back on topic, not on another poster that bowed out of the thread.

If you want to make comments like this try the pit.

I’ll hide your post.

Moderators and assumptions of good/bad faith