throwaway45644

Indeed I agree, it seems even that for some use cases training data is not anymore the bottleneck but robust test suites are. Interesting times, let’s hope we will find a responsible way to use these powerful technologies.

In theory I agree with you, we want unbiased models but here we have an input distribution that is not well understood so things get much more complex. We don’t even have a clear definition of what’s a face or not.

The model doesn’t work for people with masks: near 100% failure rate on this category of inputs. Should we release it or not?

In general some inputs are harder than others so it is expected to have more errors on those.

That being said in practice, in normal conditions, it is not hard to detect people with dark skin if the proper training data and training is used (btw, if you don’t pay attention how you do things even a low light image of a Caucasian will not be recognized) so there is little excuse to exclude a large part of the population just because of sloppiness. Moreover for this specific category (and of course others), there are consideration ethical and legal to make sure the system works for them.

Apart from that in general I do really think that ML systems with no “operator override” in many contexts are an hazard. We cannot expect the model creators to have predicted and tested for every possible input and we cannot have ways to manually correct the error (for instance in lending or border controls). Incidentally it is interesting to note this will be skilled work that will not be take over by “AI”.

There is really no reason why the dev team should include any particular demographic: how are you supposed to have 90 years old people in the team to make sure they are recognized correctly? This is a requirements issue which directly impacts validation/test data collection. If their user base has 50% black people any reasonable protocol will include enough black faces int he test data to detect the problem early on. Ml based systems will always make errors, which errors matter will be defined by market/legal/mission requirements. It may very well be that faces of black people are harder to detect (especially in backlit situations). Should you hold the product because it may not work for everybody? It’s a complex decision. Maybe you can just have a good “face detection failed” flow to handle all the errors (think not only black people but also, tattooed people, etc.).

Arguing that having quotas of that or the other in the dev team will make them more sensitive to diversity issues in general is also unnecessary because everybody is part of some minority in some situation, hence a minimum of education will make anybody understand first hand the value of inclusiveness and diversity.

Btw, the team is using only their faces to test the system they won’t go far.. (think about lighting condition / different environments).