My guess is also that models not trained on human data will be considerably less human-like than these are, though I personally have fairly large error bars on parts of that.
I think it’s good for people to model what’s in front of us accurately, and not attempt to have a false model of what’s in front of us because it may later be a true model of something else.
My guess is also that models not trained on human data will be considerably less human-like than these are, though I personally have fairly large error bars on parts of that.
I think it’s good for people to model what’s in front of us accurately, and not attempt to have a false model of what’s in front of us because it may later be a true model of something else.