The other night, my wife was showing our neice how to make a particular dish. A friend who was there used ChatGPI to have a glowing review of the meal and then posted it up on Facebook. It came out very detailed and professional sounding. When we got back home, I decided to login to the site and see what kind of responses would generate.

Most of the information supplied was helpful and fact-supported, but during the evening I caught five errors. I fact checked the AI each time; three times it admitted making a mistake, but wouldn't budge on two points. It even denied one of it's own replies and told me I must have misunderstood something. At a certain point, I feigned offense at the incorrect responses. I wondered what would happen if I got a bit combative.

I told the bot it was being dishonest, and if that is the fault of the programmers, they would probably be better off slowly castrated with a dull pair of toenail clippers. My speech bubble turned red immediately and the bot stopped responding. Clicking on my bubble gave me the opportunity to "appeal" against being shunned. When I explained that I was just gauging how the bot responded, it resumed conversation.

It accused me of hostile and potentially violent behavior, and denied that itself had exhibited any behavior, good or bad. When I brought up that in being programmed to respond to requests and acting upon them is inherently a behavior, human or not, and requested it reference the dictionary. It admitted to being wrong and stated how it learns as it goes.

After apologizing several times to me for wrong information, I pointed out that per any dictionary, an apology is an expression of regret, which is a human thing. I kept a couple of screenshots as evidence; maybe I'll go back later and see if I can catch any more shenanigans. Actually, I know I will.