The AI doesn't "feel" anything. It is simply attempting to predict the most human-like word next in a sequence of words, based off of training on millions of lines of human dialogue.
They could train the model against all of the millions of comments on hacker news, for example, and it would eventually respond to things in a way that was virtually indiscernible from the average user here.
If it says it "forgot" something, it has no actual memory. That is merely a conversational pattern/response that it picked up from common conversations it has been trained on.
They could train the model against all of the millions of comments on hacker news, for example, and it would eventually respond to things in a way that was virtually indiscernible from the average user here.
If it says it "forgot" something, it has no actual memory. That is merely a conversational pattern/response that it picked up from common conversations it has been trained on.