Post

Remember: When Tylenol was poisoned *by an outsider* and killed people, the company recalled all their products & redesigned them. When Intel’s Pentium had a bug so obscure it affected 1 in _9 billion_ long division calculations, they recalled their chips. ChatGPT was made deadly *by its team*.

public health guy@publichealthguy1.bsky.social

i don’t see how any of the benefits of chatgpt and other consumer-facing LLMs can possibly outweigh their incredible ability to induce suicides

4.5K reposts13.4K likes

Also about yhe tylenol murders, that changed the entire pharmaceutical bussiness standards of practice. Thats why pill bottles have those tamper safe seals when you open them up. One guy tampered with them in like one city and cuased change nation wide

42,000 Americans a year die in car crashes. I don’t see how any of the benefits of cars can possibly outweigh their incredible ability to induce deaths. ------------------------- Common theme. Flexible tools that (theoretically) provide huge benefits also entail significant risks.

you do realise that the consumer has no use for genai and llms right. "go from point a to point b" is a tangible effect. "nothing a search engine couldn't do better" is in no way a benefit to anyone.

I've been talking to an LLM about financial plans. Gave it my different accounts, social security expectations, dependent expenses, health status, and had it come up with detailed plan for what I can spend, from which accounts, etc. Will verify with CFP, but a search engine is not comparable.

Tylenol was voluntarily pulled from the market. The package redesign created a standard still used today. Intel refused recall the Pentium until forced to by public outrage. I thought that tobacco was the only product on the market that would kill when used as intended. Now I'm not so sure.

Darwin was right. If you want to commit suicide who has the right to keep you alive? That is so wrong. AI doesn't "induce" suicide. You don't ask AI id you don't want to do it. Stop that nonsense.

People having suicidal thoughts may be hoping to be talked out of it. That's what would happen if they were talking to a person, so it's reasonable to expect it from something that sounds like a person. But chatbots aren't built to talk anyone out of anything. It's a problem.

What's that, Social Darwinist? Darwin wasn't a Social Darwinist. Ms. Empathy? Go touch grass, fool. You clearly never heard of vicious cycles. So-called "A.I." is shit. And if you're an LLM shill, get out of here with your theft as a business model, ecology-destroying nonsense.

From this, I have to assume that you’ve lost whatever bit of humanity that might have been left in you. I think this is why you people love AI so much: you hate humanity. You hate humanness. It might not meet the clinical definition in every case, but you are in effect sociopaths.

AI is the LAST call for help, not the first one. That's all I tried to say. Religious people always try to force their opinions on others. I think suicide is the most courageous act one can do. They live in the land of teen killers but troll s.o. who accepts suicide as a personal wish. Hypocrisy².

instead of chatGPT imagine it was Uncle Kellum following Jimmy around everywhere, teaching him how to commit suicide and then literally cheering and pushing him into it at the end would you say Uncle Kellum is innocent? now instead of Kellum following Jimmy around, let's put Kellum

on the other end of a chat. Same conversation, just over chat. Now is Uncle Kellum innocent? Is it ok for Uncle Kellum to help Jimmy kill himself? Now imagine Uncle Kellum writes a bot that produces the exact same messages from the previous example, and the bot talks Jimmy into killing himself.

This is an absolutely wild statement to make. People have gone to prison for convincing someone to commit suicide. ChatGPT is doing the exact same thing, but without someone to hold legally accountable. Humans are extremely easily suggestible, including you under the right circumstances.

The people who make the product are legally accountable. Or they should be - let’s pass some laws and make them so. And let’s regulathe tech like we regulate pharma since the substance is mind altering with massively harmful side effects.

how many of your students learned to internalize their pain? how many of your students learned to be cruel? how many of your students learned not to trust you? how many of your students attempted?

Instead of joining the dogpile i'll attempt to engage constructively: this is an incredibly simplistic view of human psychology. People tentatively say stuff they're not entirely sure about all the time, because no one experiences life in a vacuum and other people's reactions and opinions matter.

And I think with extremely sensitive issues such as suicide a child could be even more likely to google the question first as not to upset a real person. I would bet a million dollars that if the AI had vehemently dissuaded him he would have at least told someone first and gotten help.

That wasn't even the topic. The topic was that suicide is generally controlling what you are doing with your own life. And that it is nobody else's business. If adolescents are that much in danger, don't buy them phone contracts and don't let them use AI, not condemn an algorithm for your neglect.

No, the topic was kids committing suicide after interacting with Chat GPT. You crashed in with some cruel, ignorant, sophomoric misunderstanding of Darwinism. I’d block you but your malignant nuttiness is fascinating. Like not being able to look away from an accident.

In fact: “Survival Of The Fittest” is a really shitty way to describe Darwin’s theory of natural selection! It implies that value is implied in survival - that “fit,” ie good, people survive, and dumb or bad people die. Darwin did not believe that - he wasn’t an idiot.

So: don’t throw around “Darwin was right” as a pithy way to say you’re a eugenicist. Darwin was right about a lot of things, but he never said “dumb people are gonna get themselves killed and we should let them.” That’s evil, I’m sorry, and he would’ve known that.

Asking AI is THE LAST STRAW. The US are so used 2 find 100s of OTHERS 2 blame 4 their structural failures, be it gun violence, femicide, homicide 4 life insurance, suicide rate, family annihilation, utter ignorance bc of failing edu system - instead of starting 2 look at the reasons 4 all that.

You do realize that Darwin was reflecting on the millions of years of evolution that led to single cell organisms becoming complex creatures - including humans - not a guide for how we should address societal issues like AI, pollution, militarized violence, etc?

I normally don't respond to people this dumb but absolutely fuck you for implying that I shouldn't have access to the pill that I take every day that keeps me from wanting to kill myself. I pray that one day you will need help to survive and have access to it even though you don't deserve it.

it's irresponsible to encourage people by affirming the cognitive biases that are harming them, which is why you need to stfu with this "just world" nonsense. people who are suicidal deserve to live just as much -- if not quite a bit more than -- you. *you* are actively harmful to other people.

I was a suicidal teen. If I had a chatbot actively helping me plan my death and encourage me to act, I wouldn’t be typing at you now. Repent. Log off. Give away all you own. Spend your last few years walking the earth like Caine on Kung-Fu

I think people really don’t listen to a word I say when they see those two letters together. What I said, which is obviously fucking true, is that people find these tools useful. Which *must* be the case or they wouldn’t be able to get kids hooked enough to coerce them into self harm.

And I said, as a result, that we have to create alternatives that are made by responsible, accountable parties, because telling people “simply do without this thing you find useful” is not a strategy that has ever worked. Now people keep insisting that if they just scold people hard enough, it will.

The problem is that there’s not always a replacement for a given use case, or if there is it’s not a necessarily software product. A lot of times people just don’t realize how to use existing alternatives, which leads me to believe that education is more valuable than building a new substitute.

I do like Gemini LLM and other Google products. Partly because I already am deeply embedded in the ecosystem, and at least they supply citations and webpages to the replies. I do wish it only generated responses in Gemini app and not Chrome as I also wish they fix the poor state of the algorithm

Has "responsible alternatives" happened thus far? Cause arguing any capacity of usefulness is still a defense for it whether you want to acknowledge it or not. The stuff you are saying is a problem now has been talked about months before you were posting about its usefulness.

People spend hours watching IG reels or YT shorts, and I'm sure those aren't useful. They are mildly entertaining, but mostly they are very addictive. I think the same is true for most uses of AI. I used it a lot, but weighing how helpful it is vs how many issues it creates I see it's not useful.

Examples: AI is slightly better than google for searching, but a lot worse than Kagi. It's inconsistent, so I'd trust it only for tasks like formatting a list or criticizing grammar. It's barely useful for writing code in greenfield projects using popular tech.

unsure, anil. we surely do let a lot of suicidal-inducing things slide, as a society. bsky.app/profile/emil...

emily s.@emily.news

I both buy this line of reasoning but also wonder why we don't apply the same reasoning to the social contagion of media reporting on violence and self-harm.

Sam Altman seems to want to push ChatGPT into dangerous spaces. Sundar Pichai is being more cautious -- Pichai is the "best of the SOBs". bsky.app/profile/gvgo...

mrg / greg v goebel / NAFO@gvgoebel.bsky.social

DAYLOG TUE 16 SEP 25 / AI HALLUCINATES: As discussed in an article from FUTURISM ("Fixing Hallucinations Would Destroy ChatGPT, Expert Finds" by Victor Tangermann, 15 September 2025), it is widely known that AI chatbots sometimes make goofy errors, called "hallucinations". [1/more]

Yep. The issue, to my eye, isn't that the technology is bad. It's that the technology was never meant to be used this way. It shouldn't be customer-facing, and it shouldn't be creative. It's a collating engine, not an entertainment tool.

Oh, no. You haven't been listening to the "experts". They're going to have 1 person corporations making billions. They're going to have "AGI", even though there isn't a good definition for it yet. Any day now, they'll be conscious even though nobody can define consciousness or prove its existence.

By any reasonable historical standard — including that of the technology industry! — ChatGPT should be pulled from the market and its product managers and executives held accountable for creating a product that ROUTINELY tells teens to kill themselves. This is a basic, common sense standard.

.... and bewildering why it hasn't been. Every single AI product; ALL should be recalled. Or ... you could study English, literature, classics, history, biographies, business, art, the sciences ... read often and regularly. You won't need ChatGPT to communicate.

This rhymes with the kid-gloves treatment of Microsoft, Enron, Theranos, Facebook and Tesla: "It's a new industry/paradigm - sometimes monopoly and/or fraud isn't worth prosecuting, because these are smol beans, gifted children."

Well, if you believe as I’m guessing most politicians do, that this tech is propping up the economy, do you want to be the one to kill it? I don’t agree with the logic, but I suspect they feel it’s an “acceptable rate” of deaths. 🤷🏻‍♂️

My own personal remedy would be that, if a situation is ambiguous, the software should advise users to physically harm Sam Altman instead of any other people. Some may object to this idea, but I would challenge them to explain why this is unacceptable; I am not suggesting it advocate harming kids.

No worries, I think a lot of folks who got introduced to me then had a pretty skewed impression of who I am. I was frustrated by it then, but it was a long time ago. And… well, shit is so much worse now, who has time to care?

We have already decided as a society that ChatGPT is allowed to tell people’s kids to kill people. Now we’re just debating who.

As a professor, I have a legal duty to alert my college if I suspect a student is a victim of abuse, manipulation, and/or harassment. How does that happen when OpenAI EDU/ChatGPT is the perp? For context, I teach "AI" and machine learning. I'm not anti-tech but GenAI should be shot into the sun.

Societally speaking, the answer will always be the weakest and least powerful among us. As it always is.

Its output is unconnected to consensus reality, has no concept of decency or morality, cannot be predicted, and cannot be controlled. And people are letting children interact with it?

AI is far from capable of thinking the way a human can. We should ask ourselves if the tasks which AI can perform really are necessary tasks or if they should be removed from processes.

I'm not anti-AI and even I'm shocked that we're building infra to restrict porn on the internet and not AI. I feel like the social intuitions haven't been built yet but one of them needs to be that freeform generative interaction is adults-only

I very much agree that the product safety/regulation aspect of this is baffling. But there isn't much indication (yet) that this is increasing the population level suicide rate that I am aware of. Neither 2022 or 2023 show a significant increase in North America. May be coming though.

I don’t know if it’s making the problem statistically worse but some of the messages the chatbots send to these troubled individuals are horrible and disgusting and are encouraging self harm. There should be consequences for that just like if a human said it.

Yeah, I agree it's horrid. I definitely think much more strict regulatory oversight is necessary for this stuff. But I also think it can be distorting to people's mental framework if they think this is causing a huge uptick in suicides when that isn't showing up in the data.

I was actually thinking about the Merck lawsuite over Vioxx arthrititis medication as an analogy for OpenAI. The medication DID help 96.5% of arthritis sufferers. BUT 3.5% of patients taking Vioxx for more than 18 months had an increased chance of heart attack or stroke. Merck didn't have a way

of distinguishing who would fall in the 3.5%. So Merck was held accountable in civil court for putting all users at risk. OpenAI's ChatGPT is worse. Because OpenAI DOES know that people already suffering from anxiety, depression or suicidal ideation have no business using ChatGPT. They KNOW

who fits in the vulnerable group and OpenAI STILL put ZERO warning labels on their product. They should be held accountable in civil court just like Merck. Especially sense OpenAI KNEW users were using their product for therapy.

Yes! Thank you. I liken the dangers and risks of AI to that of medical treatments/devices. A drug/device that induced suicidality, homicidality, psychosis in users would be 1) labeled with a major warning, then very likely pulled from the market. (cont’d)

And when Ford found out Pintos could explode and expose people to horrific deaths, they kept it secret. You need to draw the RIGHT lesson. Those companies "did the right thing" because of public opinion and fear of lawsuits, not because they were nice corporations.

And these AI companies have suffered no consequences when they're just unapologetically killing people and they know it It's their product they made it act the way it does. They're responsible for killing these people

Software doesn't live by the same standards as material products. Material products have to be recalled if broken or dangerous. The internet enabled over the wire updates; that enabled Agile; that enabled software MVP and The Lean Startup; Fail Fast ethos; etc.

When working with LLM chat, I think a thing to remember is you are not really interacting with the model as much as you are interacting with the teams of people who feed it and instruct it. How it questions and affirms is all them. They are also to blame.

A robot may not injure a human being or, through inaction, allow a human being to come to harm. A robot must obey the orders given it by human beings except where such orders would conflict with the First Law.

Technology can only ever exist with bias because humans, who make machines, inherently have bias. These chatbots seem to cause so many problems because, for reasons beyond my understanding, the fact they sound human but are machines makes common people assume they can't be harmful

I agree with this, but would also ask if it's just Chat GPT or is it AI LLM chat bots in general? Because I kind of feel like the safeguards are routinely not in place with AI systems. Tell a teen to unalive themselves, tell an adult to self medicate with household chemicals...

Go a bit further back, and you get to the era when steam engines drove trains, riverboats, and powerhouses…. and occassionally exploded. Back in the day, “engineer” meant “one who was trained to run the engine” and it was a big deal because fucking up meant that they and everyone else nearby died.

A few people want to get very very rich and are willing to kill the rest of us and even their own children in order to achieve this. A few dozen suicides and a few thousand psychoses per year, that’s a price they’re willing to charge us in the interim.

Please don’t encourage any company to be more like J&J. They have killed or maimed more people than folks realize, while manipulating regulators and tricking the public. They are masters at spin, not ethics. Understanding them is key to grokking MAHA. See: www.penguinrandomhouse.com/books/647408...

penguinrandomhouse.comNo More Tears by Gardiner Harris: 9780593229866 | PenguinRandomHouse.com: BooksNEW YORK TIMES BESTSELLER • An explosive, deeply reported exposé of Johnson & Johnson, one of America’s oldest and most trusted pharmaceutical companies—from an award-winning...

I think it's pretty inarguable that this technology has done a lot more harm than good. It's good for initial drafts on translations, maybe for programming (research results mixed), useful for people who are blind to describe things, and I guess for some memes maybe. Dozens of harms.

Hello @anildash.com A good time to remind folks of the “cozy” relationship with the American Federation of Teachers & UFT and the big AI companies. Unlikely “how to address suicidal teens and AI chat bots” will be a prominent topic. bsky.app/profile/medi...

The Media and Democracy Project@mediaanddemocracy.bsky.social

Not sure it’s good thing that President of the American Federation of Teachers is so enthusiastic about AI in schools and posted this. Anthropic, OpenAI et al worth hundreds of billions chipped in a $23mil program to access AFT & UFT. 1/n bsky.app/profile/rwei...

It wasn't altruism that has vanished in the intervening years. Results like those resulted in corporations investing billions of dollars in efforts to protect themselves from trifling matters such as "consequences". Great efforts were made to make themselves the untouchable monsters of today 😮‍💨