You're confused by the double meaning of the word "bias".

Here we mean mathematical biases.

For example, a good mathematical model will correctly tell you that people in Japan (geographical term) are more likely to be Japanese (ethnic / racial bias). That's not "objectively morally bad", but instead, it's "correct".

Western liberal culture says discriminating against one set of minorities to benefit another (affirmative action) is a good thing. What constitutes a racial and ethnic bias is not objective. And therefore Google shouldn't pretend like it is either.

> from a moral realist perspective we can still objectively judge those cultural norms to be better or worse than alternatives

No, because depending on what set of values you have, it is easy to say that one set of biases is better than another. The entire point is that it should not be Google's role to make that judgement - people should be able to do it for themselves.