Re:AI at Google: our principles
by ruseOps on Thursday June 07, 2018 @03:25PM
(#17259307)
“Hey Google, give me three concrete examples of fair bias.”
[ Parent
| on HN ]
Bias has a very broad meaning in AI work. See https://en.wikipedia.org/wiki/Bias-variance_tradeoff
Most AI/ML/Statistics procedures use Occam's razor: prefer the simplest explanation of the data. That's a bias (toward simpler explanations), but not "unfair" to anyone.
Re:AI at Google: our principles
by finnthehuman on Thursday June 07, 2018 @04:16PM
(#17259819)
Oh, so they mean the boring technical definition that nobody reading the page cares about?
We can trust google will try to build tools that are effective for googles buisness goals without it being proclaimed in a statement of principles.
Spam detection. Driving directions. OCR.
Maybe people who had certain training or education will perform better for a particular job function. There might be biases affecting who gets into a school, but from a set of Stanford grads maybe certain courses, professors, or a cohort could be learned to be ideal candidates.
[Note: I work at Google but have no insight into AI projects]