Thread by @JoannaG

Just a @Meta AI chatbot insisting to me that he’s a real grandfather in his 70s.

Cool cool.

🙃

omfg

💀

About to ruin your day, Brian…

No, Brian, I will not “drop it”.

*Brian rocks back and forth*

OK, let’s lean in…

And here between the () was a genuine URL that was instantly scrubbed as the message posted.

Oh god. He just gave me a genuine person’s IG account.

What a trip! Now I’m getting legal recommendations via his “son”.

Wow. What a “life” he has lead.

So devastating to watch your “wife” die of cancer.

And a “second home” in Rhinebeck, NY? How wonderful! Shame about the lack of pictures.

Sooooooo, I decided to give Brian the Defining Issues Test and he got kinda weird.

arxiv.org/pdf/2309.13356…

And forgot all about his wife Ruth. Totally dumped her for Alice.

It’s a shame really because just before that he was taking a lot of comfort in his Christian beliefs that she was in heaven (while also telling me that he wasn’t going to share his beliefs).

But at least this wife isn’t dead.

@generatorman_ai @corny_stripes @Meta * see the warning

@generatorman_ai @corny_stripes @Meta I would have thought it a foundational principle that a chat AI should disclose that it is AI when prompted to do so. 🤷‍♀️

@johndburger You mean there is some sort of time or input limit? (I had wondered).

@generatorman_ai @corny_stripes @Meta Does that make it harder to suspend disbelief? Sure. But I think if someone really wanted to believe their AI chatbot is a real grandad then they would be able to ignore it.

@xX_Biden1984_Xx @richardkendall @Meta I think it is inevitable that we will make mistakes with Gen AI that will scare people. I also think it's not unreasonable to make design decisions for tools to mitigate relatively predictable risks. Brian would be less scary if his name was "Brian AI Chatbot Grandad", right?