> According to slides Google shared with Motherboard—from a presentation given after the publication of that paper—its deep learning system was 87.9 percent accurate at identifying skin conditions for Black patients, the highest of any ethnicity.
It's interesting that instead of highlighting how the system is biased [1] against non-Black people, the article ignores that in favor of the bias against a different sub-category of dark skin, that is then conflated with Black [2]. At one point Google tried to improve its data set on Black people, but that didn't satisfy Vice or other journalists either [3].
[1] Using 'biased' instead of 'less accurate' is consistent with other articles, such as those on facial recognition.
[2] Tweet embedded in the article: The algorithm was developed based on training data with less than 4% dark skin types. It should come with a warning BEWARE OF RESULTS IF BLACK!!!
1 comment
[ 2.9 ms ] story [ 10.1 ms ] threadIt's interesting that instead of highlighting how the system is biased [1] against non-Black people, the article ignores that in favor of the bias against a different sub-category of dark skin, that is then conflated with Black [2]. At one point Google tried to improve its data set on Black people, but that didn't satisfy Vice or other journalists either [3].
[1] Using 'biased' instead of 'less accurate' is consistent with other articles, such as those on facial recognition.
[2] Tweet embedded in the article: The algorithm was developed based on training data with less than 4% dark skin types. It should come with a warning BEWARE OF RESULTS IF BLACK!!!
[3] https://www.vice.com/en/article/43k7yd/google-is-investigati...