1 Comment
Dec 14, 2022·edited Dec 14, 2022

While I respect that some people might have concerns about downstream consequences and (as a result) may very well decide not to publish stuff for that reason, I see it as more of a matter of personal priorities than one of ethics. If I found some way to use GPT-4chan for harm, and I did so, then the responsibility for that harm would lie solely with me. At its core, sharing an AI model is just telling other people how to use math for a certain purpose. So while acting (or, in this case, not acting) out of concern for others' interests is a noble thing to do out of personal choice, I wouldn't call it anything close to a moral duty in this case, because nobody has a positive right to not have other people know things. I realize that they *do* have a right to not have harm done to them, but merely releasing a model to the public doesn't violate this right: while one might be of the opinion that releasing the model does "a harm to society", my view is that society as a whole doesn't have rights; only individuals do. So no one's rights are actually violated until the point that a specific individual person suffers harm as a result of it, which doesn't happen merely by sharing a model alone.

Expand full comment