Pedro Domingos

It's alarming that NeurIPS papers are being rejected based on "ethics reviews". How do we guard against ideological biases in such reviews? Since when are scientific conferences in the business of policing the perceived ethics of technical papers?

US funding agencies require that procedures involving humans or animals be approved by a board that reviews their ethics. In ML, the stakes are more nebulous, but they are also potentially much higher. Why wouldn't we have some kind of ethical review process?

But that's for funding decisions, which is a very different matter. Research papers are speech, and should not be subject to censorship.

So rejecting a paper based on e.g. lack of novelty is ok but rejecting a paper for ethical issues is censorship and therefore wrong?

The point is that I'm willing to delegate the first but not the second.

Ok, but this seems to imply that the harm from blocking a paper because it is deemed not novel is greater than the harm for blocking something that is deemed unethical. That seems incredibly backwards to me.