Of all the debates and cries for LeCun's insensitivities and outrage over DeepMind's so-called "not quoting good research due to their deep racial bias", can someone comment on the scientific part? In particular, is PresGAN novel or worth quoting? Is fairness on algorithm instead of on dataset a good research direction?
don't get dragged into that false dichotomy. treating fairness as a merely technical problem is a cop out. in addition to looking at dataset and algo, look at the processes that generate data, how the data is collected/selected for the data set, and how the algo is deployed and to what end.
Can you elaborate on what you mean? I was interpreting “fairness” as “unbiased”, meaning if the process used to generate the data is biased, they are inherently unfair
