,
6 tweets,
1 min read
Read on Twitter
"The machine intelligence tested in the research, which was published in the Journal of Personality and Social Psychology and first reported in the Economist, was based on a sample of more than 35,000 facial images that men and women publicly posted on a US dating website. "
The training set is only 35,000 images, with 5 facial image per person.
Say it with me: "biased sample."
Say it with me: "biased sample."
If you look at the paper's description of how they developed the data set, there are tons of places where systematic errors can creep in.
First of all, it's for Caucasians only --- a super-critical fact that is not reported in the Guardian article.
psyarxiv.com/hv28a/
First of all, it's for Caucasians only --- a super-critical fact that is not reported in the Guardian article.
psyarxiv.com/hv28a/
Second, the data is initially taken from a data web site.
Oh, but which one? Didn't see that in the paper.
Oh, but which one? Didn't see that in the paper.
Dating web site (spelling!)
