The 'depressive ai' trigger of course will mean that the following responses will be in the same tone. As the language model will try to support what it said pr

The 'depressive ai' trigger of course will mean that the following responses will be in the same tone. As the language model will try to support what it said previously. That doesn't sound too groundbreaking until I consider that I am partially the same way. If someone puts words in my mouth in a conversation, the next responses I give will probably support those previous words. Am I a language model? It reminded me of a study that supported this - in the study, a person was told to take a quiz and then was asked about their previous answers. But the answers were changed without their knowledge. It didn't matter, though. People would take the other side to support their incorrect answer. Like a language model would. I googled for that study and for the life of me couldn't find it. But chatGPT responded right away with the (correct) name for it and (correct) supporting papers. The keyword google failed to give me was "retroactive interference". Google's results instead were all about the news and 'misinformation'.