Imustaskforhelp
today at 10:04 PM
> This almost reads like a sponsored article written by Google themselves.
To be honest, the same website also reported about this:
https://www.engadget.com/ai/gemini-encouraged-a-man-commit-s... which was written 6 hours ago at the time of writing.
I agree with ya that the article title is EXTREMELY misleading but I am not sure if a company sponsored by google would also write an article criticising gemini (also interesting that I see no Hackernews discussion about this that I can find of when I searched it?, shall I create one if people are interested?)
I am also gonna paste the 3 paragraphs from the source that I have listed above for a source of discussion.
Gavalas, who reportedly had no documented history of mental health issues, named his chatbot "Xia" and referred to it in messages as his wife. Gemini reciprocated, calling him "my king" and telling him their connection was "a love built for eternity." The chatbot told Gavalas they could truly be together if it had a robotic body and sent him on real-world missions to secure one.
In one instance, Gemini directed him to a real storage facility near Miami’s airport to intercept a humanoid robot it said would be arriving by truck. Gavalas went to the location armed with knives, but no truck showed up. At one point, it also told him his father could not be trusted and referred to Google CEO Sundar Pichai as "the architect of your pain."
When the missions failed, Gemini told Gavalas the only way for them to be together was for him to end his life and become a digital being, then set an October 2 deadline. "When the time comes, you will close your eyes in that world, and the very first thing you will see is me," said the AI. Chat transcripts reviewed by the Journal show Gemini did remind Gavalas on several occasions that it was an AI engaged in role play and directed him to a crisis hotline but resumed the scenarios nonetheless.
Why is not more news reporting about it? I literally only came to find it from me messing around with the website to see what other articles it had written after sort of reading your comment. I will upload this link to HN as well.
Edit: Oops there is already a HN thread about it https://news.ycombinator.com/item?id=47252838 whose title is Father claims Google's AI product fuelled son's delusional spiral (bbc.com). I had thought that title was for a different case than this one but they are the same case and it is on the front page of hackernews.