If you haven’t seen it yet, ImageNet Roulette is an excellent bit of educational/provocative art by the ever-brilliant Kate Crawford and Trevor Paglan. It’s a simple demonstrator of how facial recognition will label you based on its training data set. Concretely, based on ImageNet, which according to this project is the dominant training model out there, used widely in the making of real world products. I highly encourage you to try it out, but with a warning: The results might be hilarious, or pretty disturbing.
ImageNet Roulette is best enjoyed side by side with Excavating AI, which explains a little better what’s going on:
“Training sets, then, are the foundation on which contemporary machine-learning systems are built. They are central to how AI systems recognize and interpret the world. These datasets shape the epistemic boundaries governing how AI systems operate, and thus are an essential part of understanding socially significant questions about AI. But when we look at the training images widely used in computer-vision systems, we find a bedrock composed of shaky and skewed assumptions.”
I tested a few configurations of photos of myself, and depending on my facial expression had the photo labeled as a wild, and wildly wrong, range of things (including “rich person”, “grass widower”, and “econometrist”):
See also: AI based surveillance running amok. A super short “best of”:
At least seventy-five out of 176 countries globally are actively using AI technologies for surveillance purposes. This includes: smart city/safe city platforms (fifty-six countries), facial recognition systems (sixty-four countries), and smart policing (fifty-two countries).
Liberal democracies are major users of AI surveillance.
Democracies are not taking adequate steps to monitor and control the spread of sophisticated technologies linked to a range of violations.
China is the leading exporter of this type of surveillance tech, but US and European countries are in on it, too. Seriously, read this one, it’s as quick to read as it is horrifying.