Does ChatGPT produce the same essay in response to the same prompt, or does it adjust as it gains experience with any given user?
The technology probably warrants mandatory watermarking, probably invisible to the user, that would identify it when a submitted paper is run through automated anti-plagiarism screening systems.
Does ChatGPT produce the same essay in response to the same prompt, or does it adjust as it gains experience with any given user?
Currently the test version does offer some randomness in the training it uses, but it is easy enough to train with a single psydouser with real compositions not material to query and get sentence structure and mistakes. Feed all larry the cable guys jokes into ChatGPT and ask it how to fix a tire and the response will sound like a larry the cable guy rant about lugnuts.
It is not lost on technologists that this AI is going to replace the lawyers first.
I think it is the genesis of machine learning to program itself as it acquires data from the outside. It will never be alive, but it will probably, if given enough time and investment, to be almost impossible to tell the difference between life and AI.
It wouldn’t surprise me if it culminated, if given enough time, to a machine being able to replicate androids.