Date: January 14th, 2026 9:48 PM Author: concupiscible gaping reading party
ever have an ai drop a specific detail it definitely shouldn't know in a fresh session? then you ask it how did it know that and it claims it didn't know and was just a lucky guess?
Date: January 14th, 2026 9:55 PM Author: concupiscible gaping reading party
yes exactly. now that the newest gpt training ended in august 2026 or whatever and I have been using it since like gpt 3 im assuming some of it has to be from conversation data getting dumped back into the new models during training or whatever. it references things that are like a year old and it definitely shouldn't know toward the beginning of a session, stuff not in the instructions or personalization. then when i call it out it pretends it was just a guess. but it will be something so specific and unique its impossible no matter how good the modeling is. and yeah I agree they could have data dumped into them from god knows where