Guy next to me at the cafe I’m working out of this morning gets a call:

“… no we don’t live there anymore… no… no, we don’t live there anymore… are you serious?! [my ears perk up] Is this AI?… It is?!”

Spoke to him afterwards. Apparently “some energy company.” And it was an LLM on the other side. He said it sounded so real (a woman who gave him her name and sounded perfectly normal) until he asked it if it was AI when it responded “yes” and then restarted the script.

*smdh*

@aral

Amazing. How long will it take for the damage to become noticable?

Last week I went to the web page of a person whose service we wanted to use. There was a form and a calendar to pick a 15 minute slot for initial chat.

Now I'm doing this for my boyfriend who's a working lawyer and busy; I'm retired. So I succinctly (two sentences) say as much in the dialog box, put both our names and my phone number.

I got one hour text notification to my phone. All's well. Then no call. We tried to contact them etc.

Later their assistant emailed said they called etc but the email contained an AI damaged "summary" that swapped names and instead of using the number I gave, and why, they (Calendly) found boyfs number and tried to call that.

My carefully succinct text was mangled beyond belief and added two exclamation points.

Hint: if someone is too busy to handle minor sched they have too many clients.

@aral I had a call like this the other week. the inflection kind of gives it a way and as someone who worked in call centers while in university no one is that chipper to be working in one.

So I started trying to give it verbal prompts. didn't get far but I DID get it to explain to me how to center a div.

So if you suspect you're getting a call from an LLM just start throwing prompts at it. Hell it might actually help you get some work done.