Comment by tloriato

You said that the model fooled your friends 10% of the time.

I wonder how well would chatGPT | llama2 do given just the last 5 messages of each and asking to generate the next reply pre tending to be you…

Somehow I don’t think it would be worse?