Matthew M. Davis at Quillette:
Copilot seemed to follow my train of thought. I say “seemed” because I know the received opinion is that AI bots don’t have thoughts of their own and can’t really “follow” other people’s thoughts either: they just regurgitate information and predict the next word based on words they were trained on. Snicker at my naïveté if you will; I felt that Copilot was doing more than that.
It seemed to “remember” my ideas, just as it “remembered” the stories my student had given it. It could give me my ideas back in different words, and make connections among them. When I added new ideas, Copilot seemed able to take them on board and link them to things I’d already said before I could point out such connections. The “conversation” we had progressed and deepened in the way a good human-talking-to-human conversation does.
I was impressed. Copilot was doing things I had assumed generative AI bots couldn’t do.
More here.
Enjoying the content on 3QD? Help keep us going by donating now.