That's an excellent question. Another of my faves is "what are the worst consequences and how do we reduce the risk of a negative outcome? Is it possible to entirely eliminate issues responsibly?"
The former tends to stop all progress, while the latter considers that negative outcomes are real, but that we can productively reduce risk.
Kind of like how OpenAI removed racist, sexist, and other troubling images from their AI training pool and then released it in pieces (to a small group of subjects) to remove as many issues as possible. With that said, they definitely still made mistakes.
I like the motto “safety is our number three priority” but I also think that people sharing info about their danger-spraying project should take a moment to say a line or two about how they think about the issues. Instead of just ignoring them.
The former tends to stop all progress, while the latter considers that negative outcomes are real, but that we can productively reduce risk.
Kind of like how OpenAI removed racist, sexist, and other troubling images from their AI training pool and then released it in pieces (to a small group of subjects) to remove as many issues as possible. With that said, they definitely still made mistakes.