ChatGPT is pushing people towards mania, psychosis and death
Record numbers of people are turning to AI chatbots for therapy, reports Anthony Cuthbertson. But recent incidents have uncovered some deeply worrying blindspots of a technology out of control
You're viewing a single thread.
So, it’s not ChatGPT, it’s all LLMs. And the people who go through this are using AI wrong. You cannot blame the tool because some people prompt it to make them crazy.
But you can blame the overly eager way it has been made available without guidance, restriction, or regulation. The same discussion applies to social media, or tobacco, or fossil fuels: the companies didn't make anyone use it for self destruction, but they also didn't take responsibility.
First nuanced argument I've seen on this topic. Great point. Just like bottle manufacturers started the litter bug campaign. I think the problem with llm's has to do with profit-motive as well - specifically broad data sets with conflicting shit, like the water bunny next to general relativity made for broad appeal. AI gets a lot more useful when you curate it for a specific purpose. Like, I dunno. Trying influence elections or check consistency between themes.
Kitchen knife manufacturers, razor blades, self-help books, Helter Skelter, the list of things that people can “use wrong” is endless.
PEBCAK
Where is this epidemic of kitchen knives being misused?
Since when is this an epidemic?
I don't know. You're the one who brought it up, I thought you'd be the one to know.
I never used the word epidemic. I don’t believe the article use the word epidemic.
If we want to talk about things that are more damaging to People, let’s talk about social media. That is exponentially more damaging than AI.
No, you have concerns about not banning kitchen knives, razor blades, and self-help books. I wanna hear the argument.
You are creating a conversation that does not exist. I don’t have concerns about not banning kitchen knives? I’m comparing tools to tools and you are turning a conversation to something you feel like screaming and yelling about because this is the dopamine time for you.
I will not speak to someone who puts words in my mouth, and doesn’t use their ears.
If you don't want to restrict kitchen knives, why would you enter them into a discussion about "things ailing society" then? What point did you think you were making?
Kitchen knife, razor blades are a different category, for self-help books also. LLM is completely different category and there is no point of comparing knife to an llm besides to do a relativization.
The tools are relative. Pick a tool. It can be used wrong. You are special pleading, dogmatism, intellectual dishonesty.
If you’re going to refuse entire categories of tools then we are down to comparing AI to AI, which is a pointless conversation and I want no part of it.
Okay, now imagine the tool is advertised in a way that tells you to use it wrong.
"Gilette - Follow The Road, Don't Cross It™"
It's not about banning or refusing AI tools, it's about making them as safe as possible and regulating their usage.
Your argument is the equivalent of "guns don't kill people" or blaming drivers for Tesla's so-called "full self-driving" errors leading to accidents, because "full self-driving" switches itself off right before the accident, leaving the driver responsible as the one who should have paid more attention, even if there was no time left for him to react.
So what kind of regulations would be put in place to prevent people from using ai to feed their mania?
I’m open to the idea, but I think it’s such a broad concept at this point that implementation and regulation would be impossible.
If you want to go down the guns don’t kill people assumption, fine: social media kills more people and does more damage and should be shut down long before AI. 🤷♂️
Probably the same kind of guardrails that they already have - teaching LLMs to recognise patterns of potentially harmful behaviour. There's nothing impossible in that. Shutting LLMs down altogether is a straw man and extreme example fallacy, when the discussion is about regulation and guardrails.
Discussing the damage LLMs do does not, of course, in any way negate the damage that social media does. These are two different conversations. In the case of social media there's probably government regulation needed, as it's clear by now that the companies won't regulate themselves.
Okay so it has guardrails already. Make them better. Government regulations can’t be specific enough for the daily changing AI environment.
I’d say AI has a lot more self regulation than social media.
But, I run ai on bare metal at home. This isn’t chatGPT. And it will, in theory, do anything I want it to. Would you tell me that I can’t roll my own mania machine? Get out of my house lol.
Naturally the guardrails cannot cover absolutely every possible specific use case, but they can cover most of the known potentially harmful scenarios under the normal, most common circumstances. If the companies won't do it themselves, then legislation can push them to do it, for example making them liable, if their LLM does something harmful. Regulating AI is not anti-AI.
I feel the guardrails are in place, and that they will be continuously improved. If a person finds a situation where an AI suggests they kill themselves without being prompted, say, during a brainstorm about strawberry cake consistency—if you were dead you wouldn’t have this problem—would be… concerning.
If you’re going to refuse entire categories of tools then we are down to comparing AI to AI, which is a pointless conversation and I want no part of it.
The point is not to compare but analyze how AI affects us and the world around us, society. By saying "it's just a tool", or "knives can also be missuesd" you relativize discussion and that rethoric just contributes to defending openAI and other big techs and even helping them banalize the issue.
From what i witnessed is that people lose agency, get and belive fake info, everything becoming slop, people loosing jobs getting replace by more workers that are less payed etc.
EDIT: And no it's not the same as knife or razor or a gun, it will never be.
You could say the same about social media and the entire internet. Would you choose to regulate that?
I recall in the mid 90s a group of people on the street corner protesting AOL (America OnLine) and saying the internet should be stopped.
They may have had a point, but the technology wasn’t to blame for the shit that’s it’s used for.
The vague way you talk about AI makes be think that you don’t know much about it. what do you use AI for? Is it ChatGPT?
You could say the same about social media and the entire internet. Would you choose to regulate that?
Yes, fediverse is imo good(not great) model of regulation by the network(s) itself.
I recall in the mid 90s a group of people on the street corner protesting AOL (America OnLine) and saying the internet should be stopped. They were right.
The vague way you talk about AI makes be think that you don’t know much about it. what do you use AI for? Is it ChatGPT?
I use it small ammount most for translation and generating names for projects all locally.
The point is llms can steer you in harmful direction, there is much hype around it, the effects on labour, effects on human agency and knowledge etc. The power is in big tech not in some communities that use AI for their needs, but the big tech creating needs and solutions to fullfill them via advertisment, hyping it etc. That is the point of discussion not if I used it or not, or to compare it with knife or social media(that is better comparison but still) etc.
Do you think llms are that much beneficial to society or not ? What effects you see from llms in society ?
I think it would be easier to ask you why you are not using it? You’re aware of its issues, which means you won’t fall victim to them. You would only get the benefits. Have you tried the newer models of ChatGPT? They are worlds beyond GPT 3.5 or Llama 3.
I don't use online only local ones. So do you want to say how you see efects of llm on society ?
I see the effects of tools used improperly. It’s just another computer program. There’s nothing special about AI. The good and bad are way over hyped.
What does significantly more damage is social media. It’s laughable that people think LLMs are even a dent in “society.”
I can use an image manipulation program for harm, does that make it bad for society?
USERS are the problem.
It isnt exactly unheard of for regulations to be placed in the design, sale, or labeling of stuff because of misuse, to be fair. Even assuming the fault of using a tool wrong is with the user, assigning blame does not actually do anything about the problem. If enough people consistently misuse a thing in a certain way, there can be a general social benefit to trying to stop that type of misuse even if the people misusing it "are the problem", and since those people clearly arent going to just start using the thing properly just because someone pointed the finger of blame at them, addressing the problem is likely to take some kind of design or systemic change to make it more difficult for them to use that tool in that way.
I think that more effective regulation would be on social media
Found Grok.
Haha. Vote for Elon!
/s
So what is the correct usage?
It will give you whatever you want. Just like social media and google searches. The information exists.
When you tell it to give you information in a way you want to hear it, you’re just talking to yourself at that point.
People should probably avoid giving it prompts that fuel their mania. However, since mania is totally subjective and the topics range broadly, what does an actual regulation look like?
What do you use AI for?
Yeah, because someone in a manic state can definitely be trusted to carefully analyze their prompts to minimize bias.
What do you use AI for?
I don’t.
So… you have no clue at all what any of this is. Got it. I’ll bet you blame video games for violence.
Yeah, because video games are definitely the same as a search engine that tells you all your delusions are real.
Tell me more about how you’ve never ever used it and that everything you’re saying is influenced by the media and other anti-ai user comments.
Let’s see what happens when I google for UFOs or chemtrails or deep state or anti-vaccine, etc. how much user created delusional content will I find?
Lmao, I never said I’ve never touched AI I just don’t use it for anything because it doesn’t do anything useful for me.
Yes, delusional content on Google is also a problem. But you have to understand how delusional content tailored uniquely to the individual fed to them by a sycophantic chatbot is several times more dangerous.
Oh, well that explains everything, you are using it wrong.
A lot of people think that you’re supposed to argue with it and talk about things that are subjective or opinion based. But that’s ridiculous because it’s a computer program.
ChatGPT and others like it are calculators. They help you brainstorm problems. Ultimately, you are responsible for the outcome.
There’s a phrase I like to use at work when junior developers complain the outcome is not how they wanted it: shit in shit out.
So next time you use AI, perhaps consider are you feeding it shit? Because if you’re getting it, you are.
Again, I’m not fucking using it.
I played with it when it was new but it doesn’t do anything useful, I’m perfectly capable of brainstorming on my own.
Back to the topic at hand, do you not see how helping someone brainstorm their delusions with a sycophantic chatbot could be dangerous?
So today I had it do a bunch of fractional math on dimensional lumber at the hardware store. While it was doing this math for me it asked if this was for the guitar project I was working on in another chat, where I was mostly asking about magnetic polarity and various electronic, and yes it was. So then it made a different suggestion for me, which made a big impact on what I bought. I know that’s vague, but it was a long conversation.
Then, when I got home my neighbor had left a half dead plant on my stoop because I’m the neighborhood green thumb apparently. I had never seen this plant before. Took a photo, sent it to AI, and it told me what it was (yes, with sources).
Then while I was 3d modeling some shelf brackets, it helped my design by pointing out a possible weight distribution issue. While correcting that, I was able to reduce material usage by like 30%.
I don’t see any of that as “delusional”
But to the topic at hand, I think the conversations groups and pairs of humans have, both online and real life, will always be more damaging that what a single person can trick a computer into saying.
And by tricking it… you are abusing a tool designed for a different purpose. So, kitchen knives. Not meant to be murder weapons, certainly can be used for that purpose. Should we blame the knife?
I also had it make you this image:
