| homepage | https://yongyeol.com |
| https://twitter.com/yy |
Following #fediverse instance data analysis by @estebanmoro (see https://mastodon.social/@estebanmoro/109331650343049058), here are the mandatory <s>power-law</s> 👀 distribution plots.
As one may expect, the size of the instances shows heavy-tailed distributions for users and activities. I also assume that low-activity instances would have been under-sampled.
Bianchi et al., "Easily Accessible Text-to-Image Generation Amplifies Demographic Stereotypes at Large Scale" https://arxiv.org/abs/2211.03759
ML models tend to not only *reflect* but also *amplify* social biases.
Most housekeepers are white, but the models generates almost no white housekeepers. Only about half of the software engineers are white, but most generated images of software engineers are white.
Machine learning models are now able to convert user-written text descriptions into naturalistic images. These models are available to anyone online and are being used to generate millions of images a day. We investigate these models and find that they amplify dangerous and complex stereotypes. Moreover, we find that the amplified stereotypes are difficult to predict and not easily mitigated by users or model owners. The extent to which these image-generation models perpetuate and amplify stereotypes and their mass deployment is cause for serious concern.
As the "parody" becomes an issue... The Onion filed an amicus brief the other day to the US supreme court. It may be the most entertaining legal opinion ever in history (and the only legal opinion I have ever read :P)
It's amazing.
