3 ms·

> Men have long dominated fields such as finance and science, for example, so the technology would conclude from the data that an investor or engineer is “he” o

> Men have long dominated fields such as finance and science, for example, so the technology would conclude from the data that an investor or engineer is “he” or “him.” Err, so the AI correctly predicts that statistically you probably intend "him", but we limit the utility of the tool because that would be discriminatory? I know it will get it wrong sometimes, and you can say it re-inforces stereotypes, but if it will get it right most of the time based on strict statistical inference, seems like it could at least be configured. It seems to be a case where the AI is a bit too accurate... I don't disagree with their decision, I think given the circumstances it's actually a brilliantly safe move and keeps out of the fray as much as possible. > “The only reliable technique we have is to be conservative,” I know this is just whimsical, and a horrible logical equivocation on my part, but that's kind of funny that a large tech company decided being conservative might actually be useful in some cases to help protect against liberal outrage...

cix_pkez 8y ago
And I find it ironic that in this discussion of choosing words carefully, we still use the basically meaningless (because of all of the different ways to use them) words "conservative" and "liberal". The words have nonpolitical meanings, and political meanings. And even the political meanings are inconsistent.
ncr100 8y ago
A more product quality reason - besides political - is the implicit bias corrupts intent.
damnyou 8y ago
Yes, the goal of the modern Civil Rights Movement has basically been to say "in these areas, even though Bayesian inference produces statistically valid results, we as a society have decided that it's immoral to use it." What axes are OK to use Bayesian inference on and what are not is a philosophical and historical question with lots of practical implications across politics, economics, actuarial science etc. It's really worth thinking about. But here's a start: in general, people are more forgiving of inferring data from mutable traits than immutable ones.

cix_pkez 8y ago

And I find it ironic that in this discussion of choosing words carefully, we still use the basically meaningless (because of all of the different ways to use them) words "conservative" and "liberal". The words have nonpolitical meanings, and political meanings. And even the political meanings are inconsistent.

ncr100 8y ago

A more product quality reason - besides political - is the implicit bias corrupts intent.

damnyou 8y ago

Yes, the goal of the modern Civil Rights Movement has basically been to say "in these areas, even though Bayesian inference produces statistically valid results, we as a society have decided that it's immoral to use it." What axes are OK to use Bayesian inference on and what are not is a philosophical and historical question with lots of practical implications across politics, economics, actuarial science etc. It's really worth thinking about. But here's a start: in general, people are more forgiving of inferring data from mutable traits than immutable ones.