core-questions
Joined 8 years ago
Joined 8 years ago
It seems it's best to think of "cold" and "flu" as totally informal terms with a wide range of possible meanings. I've had Actual Real Influenza twice and both times it knocked me on my ass for a week, lost weight, etc. yet I hear people saying "I have the flu" when they have mild symptoms and only miss a single day of work. Hard to take such a wide range medically seriously.
My career was made by someone finding my resume in a trash can.
https://geekfolk.blogspot.com/2016/02/remembering-dos-shell....
Seems like this logic doesn't apply to payment processors, web hosts, and anyone else who doesn't want to deal with free speech that they find offensive. Bake the cake, right?
Not in the slightest, no. I prefer the average IQ of the world to not drop by a standard deviation - you think we'll get out of climate trouble when most people are functionally illiterate?
Well, we might, in the sense that we might lose the ability to manufacture things that produce significant carbon.....
Doesn't work like that. If people who retain the values and views that formed the West (e.g. deep inquiry into all topics regardless of orthodoxy) abdicate all responsibility for weighing in when we see these values being attacked, how exactly is that going to help the cause?
> Should we stop studying all forms of bias in computing?
Loaded question. Studying how ML systems reflect data is fine; but when ML systems give us results that accurately reflect real world data, but that we don't like based on our progressivist / ultra-liberal priors, there's a problem.
Statements an ML system can trivially tease out of a data set, like "women are more likely to do career X" or "this group is more likely to commit murder" or "this person is more likely to be hired for this position". When this reveals an uncomfortable truth that has been pre-ordained as offensive regardless of its basis in reality, we have a choice between rejecting the facts based on our priors, or updating our thinking to reflect reality. Sadly, the trend has been to do the former, and to punish anyone doing the latter.
Just go look at what Amazon had to do with their ML-based resume sorting system. It revealed that the best resumes (sorted by likeliness to be hired) were of Asian and White presenting males; which we all know to be true for their hiring, because we all know that that is the current demographics of the industry in general. Of course, such a system cannot be allowed to operate, even though the humans which continue to do the screening in its place will continue to apply the same selection rubric they have before. Nothing is solved, nothing is learned. It just gets chalked up to "systemic bias" without actually investigating any of the reasons for this (such as IQ distributions, access to technology, and whatever raw factors dictate people's individual and aggregate interests toward career selection).