That is an interesting concept to toy with user expectations. I don’t know how well it would be received but I’d love to see data from such an experimentation.
I wouldn’t, it’s not going to be meaningful after one or two tries.
I suppose it could be interesting if it was announced in advance that Eliezer was going to try it and then we could spend the next few months accusing each other of being Eliezer witch-hunt style, except with Bayesian priors. Seriously, I am in favor of doing it that way.
Perhaps, contact someone likely and ask them to paraphrase the post in their words and submit it as their own?
Now we’ll be getting all kinds of posts with, “Eliezer did not write this..or maybe he did!” …
That is an interesting concept to toy with user expectations. I don’t know how well it would be received but I’d love to see data from such an experimentation.
I wouldn’t, it’s not going to be meaningful after one or two tries.
I suppose it could be interesting if it was announced in advance that Eliezer was going to try it and then we could spend the next few months accusing each other of being Eliezer witch-hunt style, except with Bayesian priors. Seriously, I am in favor of doing it that way.
Thinking of the recent art threadjack I was party to recently—I wish the art community would do that! Without the paraphrasing, though.
And then, of course, be sure to get independent evaluations of a work before discussing it with anyone to prevent information cascades.