[belval]

Ethics in AI are already an ill-defined and gigantic field, and while I agree with your point on diversity, Google is a business that wants to create value. How do you create value with ethics? By positioning your product as being less biased than the next.

There is a lot (really a lot!) of discoveries that are needed to identify and "fix" biases in DL models. The current method actually used is to test the obvious. i.e. "does our model work as well on all skin colors" which is far from being a robust and systematic method.

That's what Google actually needs and there is still a lot of fundamental research to do before getting there. Having ethics researchers writing papers on BERT generating too much CO2 is not only something an accountant could tell you, it's also not actionable as they are already trying to use zero-carbon electricity. Having researchers focus on inequalities in the US is good but it's again not actionable without the fundamental methods that need to be developed.

TLDR: Google needs researchers that are trying to solve the hard problems of the DL algorithms, not that get side-tracked by problems that are a matter of public policy.