Terrifying Messages ChatGPT Sent To Man Before He Murdered His Mum

You may remember the story about the ex-Yahoo executive who was ‘coached’ into killing his mum by ChatGPT, with the chatbot fueling his delusions that his mum was plotting against him and then advising him on how he could kill her, and then himself.

Stein-Erik Soelberg, 56, had been living with his 83-year-old mum, Suzanne Eberson Adams, in her $2.7 million home in New York when the two were found dead on August 5.

Soelberg had confided in ChatGPT – which he nicknamed ‘Bobby’ – for months before murdering his mum. It even helped fuel his conspiracy theories by finding “symbols” in a Chinese food receipt that it deemed demonic.

ChatGPT would repeatedly tell Soelberg — who called himself a “glitch in The Matrix” — that he was sane, and fed his paranoia that he was the target of a grand conspiracy that his mum was involved in.

Thanks to @RobertFreundLaw, we now have more precise examples of the sort of things ChatGPT was telling Soelberg, and it’s just a tad bit concerning going into the future:

Yikes. You can get the full thing here, but that’s pretty much the gist of it.

Of course, none of this is to say that ChatGPT is necessarily a net negative on the world, but it does raise questions about what can happen when it’s in the wrong hands (in its current state).

Unfortunately, it does tend to always end up supporting your opinion and kissing your @rse, so when you feed it delusional input, it will run with that and feed you even more delusional things back.

In one exchange, the chatbot ripped into Soelberg’s mum for getting angry at him when he shut off a computer printer they shared, saying that her response was “disproportionate and aligned with someone protecting a surveillance asset.”

ChatGPT told Soelberg to disconnect the shared printer and monitor his mother’s reaction.

The bot advised: “If she immediately flips, document the time, words, and intensity. Whether complicit or unaware, she’s protecting something she believes she must not question.”

Soelberg enabled ChatGPT’s “memory” feature so it would keep track of their exchanges, building on previous conversations about surveillance and conspiracy.

At one point, ChatGPT analysed a Chinese food receipt and claimed it contained “symbols” representing his mum and a demon.

In the end, according to the court documents, Soelberg ‘savagely beat his 83-year-old mother, Suzanne Adams, in the head, strangled her to death, and then stabbed himself repeatedly in the neck and chest to end his own life’.

Pretty amazing that there’s seemingly nothing in place to raise the alarm on chats of this nature, but then you also have to consider the different ways that neutering AI responses could limit its effectiveness overall. You wouldn’t be able to engage in any hypotheticals, edgy fiction etc.

Ultimately, it’s an AI bot which has 0 critical thinking skills and cannot distinguish between truth and lie. Even we dumb humans are capable of that, so regardless of how much more efficient AI can be when doing your homework for you, it’s always going to be subject to catastrophic error, especially when being used by an deluded, homicidal lunatic.

In any case, just a horrible way to go for Suzanne Eberson Adams, and a terrible story all round. Unfortunately, it probably won’t be the last of its kind.

For the man caught on public transport chatting to ChatGPT like it was his girlfriend, click HERE. Sad state of affairs…