From the course: AI Accountability: Build Responsible and Transparent Systems (2022)
Unlock the full course today
Join today to access over 24,900 courses taught by industry experts.
Biased labeling of data
From the course: AI Accountability: Build Responsible and Transparent Systems (2022)
Biased labeling of data
- [Instructor] A couple years ago, I was shocked to learn that indoor rowing or rowing on an erg wasn't just something people on crew did to train in the off season but it's a competitive event all on its own. Now, some people love it and I am now one of them but other people think it's the craziest thing they've ever heard. Similarly, a lot of people really like scary movies, like my children, but I have to close my eyes and hide even during the trailers for them. I, for one, am not the intended audience for horror. It reminds me of a Latin proverb. (instructor speaking in Latin) Or there's no accounting for taste. It refer to the idea that people's preferences can sometimes vary in unpredictable ways, to each their own, and that's great when you're talking about personal inclinations. But when you're trying to develop a machine learning algorithm that can serve a large number of people, you have some major challenges.…
Practice while you learn with exercise files
Download the files the instructor uses to teach the course. Follow along and learn by watching, listening and practicing.
Contents
-
-
-
-
The challenge of classification errors3m 39s
-
(Locked)
The causes of classification errors6m 18s
-
(Locked)
Bias in AI3m 50s
-
(Locked)
Supervised and unsupervised learning8m 16s
-
(Locked)
Biased labeling of data7m 15s
-
(Locked)
Construct validity6m 14s
-
(Locked)
The absence of meaning4m 54s
-
(Locked)
Vulnerability to attacks4m 45s
-
-
-
-
-
-