The article explicitly recommends building systems which can't learn those things, and suggests characterizing them is a first step:

"We recommend addressing this through the explicit characterization of acceptable behavior. One such approach is seen in the nascent field of fairness in machine learning, which specifies and enforces mathematical formulations of nondiscrimination in decision-making (19, 20). Another approach can be found in modular AI architectures, such as cognitive systems, in which implicit learning of statistical regularities can be compartmentalized and augmented with explicit instruction of rules of appropriate conduct (21, 22)."

According to the article, the most closely related work "is concurrent work by Bolukbasi et al. (6), who propose a method to “debias” word embeddings."

(Recall that in the context of the article, "bias" is something that generates true predictions that are objectionable rather than something which is false.)

- An AI correctly infers (simply by reading text) that a physicist is male and a nurse is female.

- An AI correctly infers the gender of humans with androgyonous names.

- An AI infers insects are unpleasant and flowers are pleasant to humans.

- An AI also infers that African American names are more likely to be associated with unpleasantness than European names.

[edit: to those who dislike this comment, can you tell me what you object to? Which of my concrete examples is not in the paper?]

In any case, the paper suggests that this "bias" or "prejudice" is better described as "truths I don't like". I'm asking if the author knows of any cases where they are actually not truthful. The paper does not suggest any, but maybe there are some?

Near as I can tell, your paper shows that these "biases" result in significantly more accurate predictions. For example, Fig 1 shows that a machine trained on human language can accurately predict the % female of many professions. Fig 2 shows the machine can accurately predict the gender of humans.

Normally I'd expect a "bias" to result in wrong predictions - but in this case (due to an unusual redefinition of "bias") the exact opposite seems to occur.

(Drawing on your analogy with stereotypes, it's probably also worth linking to a pointer on stereotype accuracy: http://emilkirkegaard.dk/en/wp-content/uploads/Jussim-et-al-... http://spsp.org/blog/stereotype-accuracy-response )

The Nazi in question did not engage in any violent acts at that gathering. That makes him a peaceful protester.

It's the rare subversive academics, e.g. Charles Murray, who are being assaulted and chased of campus for holding unorthodox opinions.

And it's happening off campus as well - for instance, we recently had a peaceful (but unorthodox) anti-war activist assaulted: https://antifascistnews.net/2017/04/09/richard-spencer-just-...

On the contrary, most of us would prefer this. That's why yellow cabs are losing market share everywhere that men with guns don't take away their right to choose.

In a taxi you can rip off a firangi or fail to pick up a black passenger at your leisure. With Uber, this will cause your ratings to drop below 4.3, or your acceptance rate/cancellation rate to drop below/above whatever threshold they use. Then you get kicked off the platform.

Uber is providing the necessary regulation of the system that governments fail to provide.

As a consumer, what exactly should I be angry about?

Racism isn't like homeopathy, where any quantity at all has the same effect. More is worse, less is better.

In that study, the median black taxi rider is passed by 2 taxis before being picked up, vs 0 for the median white (see fig A.6). The difference for Uber is not remotely as large.

Some folks may hate them because of Susan Fowler, but that doesn't change the fact that they've drastically improved the world for consumers.