> So, the unwanted results here are still due to the input data. But neither the input data, nor the algorithms contain any racial bias towards the black people. The results might be merely due to the difference between dark faces and pale faces under the natural law of physics. These are unwanted results due to our social norms, but they are NOT racially biased or racist, because there is no such bias introduced or inherent during any part of the process.

In this, and the reasoning above it, I think you are correct. Assuming we are right about this, what is the next step? You say the next step is to improve data collection, e.g. by creating better cameras. That seems a fine proposal to me, I support that, but I think there is more that might also be done. For instance, the use of ML models could be restricted or regulated in at least some contexts, until the problems with data collection are rectified. For instance, we could ban the police from using facial recognition models until the problems with data collection are solved. The bias is a side-effect of photography, not something intrinsic to the facial recognition algorithm, but restricting the use of that algorithm might nonetheless be a valid response to this circumstance.