,
6 tweets,
2 min read
Read on Twitter
A failure of "human oversight" over the Facebook algorithm by @MikeIsaac is one example of *many* to come. 1/ nyti.ms/2yes6gu
In Facebook's case, the harm was ad-targeting based on racist views. But the broader-human in the loop-problem will crop up everywhere. 2/
What about when the harm isn't hate driven advertising, but algorithms that target suspicious behavior? Or a trespasser on your lawn? 3/
The "whoops" moment there will be more disastrous. The after the fact "failure of human oversight" isn't going to work. 4/
We're not designing in more Stansilav Petrovs into the world. Maybe we should be. end/ nytimes.com/2017/09/18/wor…
[And duh I don't mean more Soviet officers, I mean more human beings who can spot errors.]
