Excited to launch this new investigation w - to be read alongside ImageNet Roulette. It's about how training data works, the costs of classification, and why "removing bias" or "increasing diversity of data" isn't enough 👁️
Startled by the racist, sexist, odd results of ImageNet Roulette? Wondering how AI learns to see, what the politics of its perception are, and who decides? The brilliant and break it down
Whoa. Truly honored that , a multi-year investigation with into AI training sets like ImageNet, is the joint winner of the Ayrton Prize from the British Society for the History of Science
Computer vision tools are being used in many questionable ways and are often built on unjustified assumptions. Given this reality, the defiant view that "speech and images are solved problems" might make it hard to hear the "yelling" about bias.
This piece by & is probably the most incisive & impt piece of writing I've read on the subject of ethics/politics in data science. Please read & share -- "Excavating AI: The Politics of Training Sets for Machine Learning"!
This is certainly not a ‘fix’, and there are still half a million people’s photos there without their knowledge or consent, classified in ways that they’d likely reject. Also, deleting this history raises its own problems- as we explain here
Required Reading! Excavating AI: The Politics of Images in Machine Learning Training Sets
'What if the challenge of getting computers to “describe what they see” will always be a problem? In this essay, we will explore why the automated interpretation of images is an inherently social and political project, rather than a purely technical one.'
We're all interdependent here as brilliantly showed with
"The Politics of Images in Machine Learning Training Sets” A very interesting and important read from &
A long, very interesting and rather important, read on The Politics of Images in Machine Learning Training Sets
While there are deep social problems reflected in some of the included images and the WordNet taxonomies it was based on (famously discussed by & : ), this does not reduce the ImageNet project in its entirety to an act of hubris.
Absolutely wild that a paper and response launched by and ImageNet Roulette makes absolutely no mention to those projects or their authors