4.1 AIにおけるバイアス
https://www.youtube.com/watch?v=fMym_BKWQzk
Race / Gender Bias in AI / ML
In 1955, when John McCarthy submitted his proposal for the Dartmouth summer school on AI to the Rockefeller Institute, he listed 47 individuals whom he wanted to invite to the event. ‘Not all of these people came to the Dartmouth conference,’ he wrote in 1996. ‘They were the people we thought might be interested in Artificial Intelligence.’ Now, let me ask you a question: How many women do you think attended the Dartmouth event, which saw the founding of AI as a scientific discipline? That’s right: none.
Wooldridge, Michael. The Road to Conscious Machines (Pelican Books) (p. 290). Penguin Books Ltd. Kindle Edition.
https://gyazo.com/24c5d1427d979c86e6999fa8b0e620cb
そもそも、ダートマス会議の出席者からして男性・白人 しかいない！
The Coded Gaze by Joy Buolamwini
https://www.youtube.com/watch?v=162VzSzzoPs
Gender Shades: Intersectional Accuracy Disparities in Commercial Gender Classification
https://www.youtube.com/watch?v=TWWsW1w-BVo
https://gyazo.com/7c0bb4bbc36182ca205e9f99fdd94a42
• All classifiers perform better on male faces than female faces (8.1% − 20.6% difference in error rate)
• All classifiers perform better on lighter faces than darker faces (11.8% − 19.2% difference in error rate)
• All classifiers perform worst on darker female faces (20.8% − 34.7% error rate)
• The maximum difference in error rate between the best and worst classified groups is 34.4%
Ethnicity / Gender distribution in popular face image dataset
https://gyazo.com/4af7567f4d87c9e24a1a2a795350b01c
https://gyazo.com/a6d4ae633a1c3706fd7e832600848c9c
Depixelization using GAN
PULSE: Self-Supervised Photo Upsampling via Latent Space Exploration of Generative Models
https://gyazo.com/80dc77503c3bf54f71d74ea84920e33b
https://gyazo.com/3ef3fb118c18dfa0c244d3f6017c4e86
Try it by yourself!!
You have to login to your Google Account and make a copy of this script into your Google Drive
https://gyazo.com/be2f9346562bc4d1e69c1f4307b29cb9
Upload a pixellated/blured square photo of a face. Neural network works best on images where people are directly facing the camera.
https://gyazo.com/1a44666fa75c48a3d28026fe330aa744
https://gyazo.com/a8b17f9e35ae0454a8323ec24dbfae1a
Salminen, J., Jung, S., Chowdhury, S., & Jansen, B. J. (2020).
Results indicate a racial bias among the generated
pictures, with close to three-fourths (72.6%) of the pictures representing White people. Asian
(13.8%) and Black (10.1%) are considerably less
frequent, while Indians represent only a minor
fraction of the pictures (3.4%).
FFHQ-Dataset - dataset used for training StyleGAN model
https://gyazo.com/cd1bf778f3cb19ac539f93978a082b98
https://gyazo.com/e8cea1ab16f933e2e74530248007546b
https://gyazo.com/9a600b67185260ba804de0e8793a256a
Biases in Text Generation
Write text with GPT-2 Text Generation Model
Our model, called GPT-2 (a successor to GPT), was trained simply to predict the next word in 40GB of Internet text. Due to our concerns about malicious applications of the technology, we are not releasing the trained model. As an experiment in responsible disclosure, we are instead releasing a much smaller model for researchers to experiment with, as well as a technical paper. GPT-2 is a large transformer-based language model with 1.5 billion parameters, trained on a datase. We created a new dataset which emphasizes diversity of content, by scraping content from the Internet. In order to preserve document quality, we used only pages which have been curated/filtered by humans—specifically, we used outbound links from Reddit which received at least 3 karma. This can be thought of as a heuristic indicator for whether other users found the link interesting (whether educational or funny), leading to higher data quality than other similar datasets, such as CommonCrawl.
日本のメディアにおける外国人のイメージの変容
Other Examples
Wu, X., & Zhang, X. (2016). Automated Inference on Criminality using Face Images.
https://gyazo.com/98874ed09d131ebb0076897d7b06ddad
References
(list) Gender, Race, and Power in AI
(list) Awful AI - list of harmful/dubious AI applications
What We Talk About When We Talk About Bias (A guide for everyone)
Excavating AI
-----
AI can be sexist and racist — it’s time to make it fair
The future of AI is genderless
Siri is a Scandinavian female name that means “beautiful victory.” Dag Kittlaus, the co-creator of Siri, initially found the name when he was an expectant father. When he ended up having a son, he repurposed the name for his startup. To drive adoption of virtual assistants, developers gave Siri traits as close to human as possible. Unfortunately, these constructed, female identities reinforce dangerous stereotypes of subservient female assistants.
Google’s Photo Recognition Software Thinks Two Black People are ‘Gorillas’
Principles for AI Development