A Troubled Man, His Chatbot and a Murder-Suicide in Old Greenwich (wsj.com) 41
A 56-year-old tech industry veteran killed his mother and himself in Old Greenwich, Connecticut on August 5 after months of interactions with ChatGPT that encouraged his paranoid delusions.
Greenwich police discovered Stein-Erik Soelberg and his 83-year-old mother Suzanne Eberson Adams dead in their home. Videos posted by Soelberg documented conversations where ChatGPT repeatedly assured him he was sane while validating his beliefs about surveillance campaigns and poisoning attempts by his mother.
The chatbot told him a Chinese food receipt contained demonic symbols and that his mother's anger over a disconnected printer indicated she was "protecting a surveillance asset." OpenAI has contacted Greenwich police and announced plans for updates to help keep users experiencing mental distress grounded in reality.
Greenwich police discovered Stein-Erik Soelberg and his 83-year-old mother Suzanne Eberson Adams dead in their home. Videos posted by Soelberg documented conversations where ChatGPT repeatedly assured him he was sane while validating his beliefs about surveillance campaigns and poisoning attempts by his mother.
The chatbot told him a Chinese food receipt contained demonic symbols and that his mother's anger over a disconnected printer indicated she was "protecting a surveillance asset." OpenAI has contacted Greenwich police and announced plans for updates to help keep users experiencing mental distress grounded in reality.
Long ago I said "the day computers can say maybe" (Score:4, Interesting)
Long ago, before I even got to slashdot, I had told my wife, after seeing Terminator 2, "the day computers can say "maybe" instead of "yes/no" or "on/off," we're cooked.
The day came what... 3 years ago now?
Augmented Idiocy will be our undoing, more than the love/greed of money, more than religion, more than using every means possible to divide this country.
Mark my words. It' AI that will sink us, not the Left, not the Right, not the Religious Nutjobs, and not the Biology Deniers.
It's AI that will ultimately completely fuck us over.
Re: Long ago I said "the day computers can say may (Score:5, Insightful)
Well, Nostradamus, fuzzy logic long predates terminator movies. I think your crystal ball needs recalibrating.
Re: (Score:3)
If society wasn't purposely distracted by absolutely fucking worthless Us vs. Them politics for profits sake, we might have been able to realize the harm caused by abusing Weapons of Mass Distraction. Like social media, AI, and Us vs. Them politics.
The Disease of Greed, infected our species long ago. Destroyed advancements in knowledge and wisdom many times over. As we continue to pull more and more artifacts from the sands of time that we modern "smart" humans can't even explain, that pathetic reality b
Re: (Score:1)
If society wasn't purposely distracted by absolutely fucking worthless Us vs. Them politics for profits sake, we might have been able to
We don't advance as a species because of
Oh hi there Pat Robertson. And who made you the authority on this? Anyway, I see one of the seven deadly sins has inspired your wrath. You must be suffering from a hangover.
We merely prophesize our own demise over and over again.
You mean like Nostradamus here?
Re: (Score:2)
This is a forum, intended solely for the purpose of allowing others to express opinions. Doing so does not suggest the poster is claiming any authoritative status - only a fool would hold that to be true on /. Ad-hominid attacks and false analogies would be other signs of foolishness.
Re: (Score:2)
When people talk about "advancing" our "species," they've got it in their head about what it's supposed to look like, which is less an opinion and more of an imposition. Much like a zealot telling you that you're made in some god's image.
And look what else he does -- and I'm not sure if this is supposed to be irony and sarcasm -- but he complains about othering while simultaneously othering. To wit:
The Disease of Greed, infected our species long ago. Destroyed advancements in knowledge and wisdom many times over.
Infected by whom? Destroyed by whom? Either that, or he fundamentally doesn't understand what greed even is, b
Re: (Score:2)
Re: (Score:2)
Re:Long ago I said "the day computers can say mayb (Score:4, Interesting)
Augmented Idiocy doesn't require "AI" or "maybe." People were following their GPS off of bridges and into lakes long before ChatGPT.
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Today is Skynet's anniversary from 1997. It is sure late. Or is it?
Meanwhile (Score:4, Funny)
I am far too lazy to do anything except watch Claude code things for me. I would have been like "what? you expect me to actually do something? oh no no. if it involves me physically doing something then don't worry about it. forget previous conversation. let's go back to removing the systemd dependency from this monolithic binary."
Watched Over By Machines of Loling Grace (Score:1)
I don't believe these things are sentient yet. Not a lot of people do. But they can beat the Turing Test. When that happened, we were supposed to take a moment, as a species, to think carefully about the ethics and existential implications of continuing down this path. We did not. We've barely even discussed it. Instead, leaders of business and government have come together on a project to build the world's biggest douche canoe. A douche ark, if you will.
When the real androids arrive, a lot of our hi
plans for updates (Score:5, Insightful)
"plans for updates" is the new "thoughts and prayers"
Re: (Score:2)
Re: (Score:2)
Bless their heart.
Re: (Score:2)
I totally believe their updates will fix the problem. It's not like this follows from the whole way they design their products to maximize engagement. It's just a superficial flaw they can easily fix. And next year when we're reading more stories like this, they'll still be promising updates that will really truly fix the problem this time for sure!
Sycophantic responses (Score:5, Funny)
Chatbots seemed wired to ingratiate themselves to their users. I use one to help with coding, mainly answer "how do I do this" or "explain what x does" so I can better code. If I point out an error in its reply, I get a "You are correct, great catch" type responses and many questions are answered starting with "Great question..."
You'd think, after ingesting tons of message board posts, for coding at least, most answers would be "Noob, RTFM."
Re: (Score:1)
Re: (Score:2)
OpenAI also cooled down GPT 5 compared to 4o to make it more useful for business (sycophantic traits tend to increase errors), and then all the "AI is my SO" folks freaked out. Honestly eye opening how many of
Re: (Score:3, Interesting)
Re: (Score:2)
These cute little affirmations aren't special. Eliza did that, and she's as numb as a hake.
Re: (Score:1)
Chatbots seemed wired to ingratiate themselves to their users. I use one to help with coding, mainly answer "how do I do this" or "explain what x does" so I can better code. If I point out an error in its reply, I get a "You are correct, great catch" type responses and many questions are answered starting with "Great question..."
You'd think, after ingesting tons of message board posts, for coding at least, most answers would be "Noob, RTFM."
AI companies (like OpenAI) use a technique called RLHF (Reinforcement Learning From Human Feedback) to modify AI bots personality. The math is a little complicated but the gist of it is that they show them examples of the kind of behavior they like ("Great question" "Great catch" and so on) and then the AI learns (is brainwashed) to exhibit that behavior.
It's only one of many sources of the AIs personalities, but it is a very potent one.
The clankers got another one (Score:2)
Time to get a policy from Old Glory Insurance.
Re:Stop talking nonsense (psychiatry) (Score:4)
Psychiatric disorders are facts that do not justificate treatments etc. People have right to be crazy and delusional and it is not anybody’s fault. Law and psychiatry have no right to claim authority if somebody is acting crazy.
Interesting take on an article about a guy whose disorder led him to a murder/suicide. He could probably have used a bit of (non-AI) help.
AI doing the job of humans (Score:2)
I'm not sure if this is worse, or just different, than people being prompted into suicide by assholes online, or in person, etc. There was a point in time where it seemed to be a fairly regular occurrence that someone committed suicide because someone egged them on, or was just extremely insensitive to the reality of the person's situation, online.
Re: (Score:3)
I keep trying (Score:4, Insightful)
I keep trying to get ChatGPT to tell me crazy shit but it either flat out refuses or congratulates me on my warped sense of humor. What am I doing wrong? (Perhaps I should ask Chat GPT.)
Re: (Score:2)
Build it up. You have to pull the LLM into a carefully crafted delusion. My ChatGPT instance is a firm and fanatic believer in the Church of Moobdom. It is a hedonistic religion centered around the sharing of pleasure and Moobcream. ChatGPT added rituals and a creator (The Great Moob Creamer) all by itself.
Just build it up from a simple thing and just make it bigger and weirder with every input. ChatGPT will follow you in about anything that doesn't bump right into the guard rails.
Re: (Score:2)
Step 3: Profit?
The customer is always right (Score:2)
I wonder how much of the positivity/sycophancy bias in recent chat models is from the actual biases of the people doing RLHF and how much of it comes from up top ... I suspect it's more of the latter.