They also supposedly added the mask recognition "cut-off" because the ML was mistakenly learning people's masks and you could impersonate some people by wearing

They also supposedly added the mask recognition "cut-off" because the ML was mistakenly learning people's masks and you could impersonate some people by wearing the same mask as they regularly wore. Another case where statistic models were likely to learn the wrong things.