AI Model Identifies Sexual Orientation - In-depth Analysis

This is a past event

22 people went

Every week on Tuesday

Algolia

55 Rue d'Amsterdam · Paris

How to find us

We meet in the lobby. Ask reception for location of group.

Location image of event venue

Details

We'll prepare for this evening's AI Town Hall Meeting (https://www.meetup.com/Paris-Machine-Learning-Study-Group-in-English-Meetup/events/261892701) by performing an in-depth scientific analyses of tonight's papers (primary and replication papers). The research claims to have created AI capable of identifying sexual orientation (gaydar) from a single image.

Tuesdays are applied machine learning day. We have a peer-to-peer discussion with a focus on an applied machine learning topic. We also meet on Fridays when we discuss a predetermined research paper.

Abstract: We show that faces contain much more information about sexual orientation than can be perceived and interpreted by the human brain. ... Those findings advance our understanding of the origins of sexual orientation and the limits of human perception. Additionally, given that companies and governments are increasingly using computer vision algorithms to detect people’s intimate traits, our findings expose a threat to the privacy and safety of gay men and women. A heat map displaying clearly defined facial differences would be a major finding.

Primary Papers:

1. Deep neural networks are more accurate than humans at detecting sexual orientation from facial images: https://psyarxiv.com/hv28a/

2. Author's Notes: https://docs.google.com/document/d/11oGZ1Ke3wK9E3BtOFfGfUQuuaSMR8AO2WfWH3aVke6U

3. A Replication Study: Machine Learning Models Are Capable of Predicting Sexual Orientation From Facial Images: https://arxiv.org/abs/1902.10739

Discussion:

1. I'm very surprised that I can't any research which detects handedness from images. What prompted me to find such research is Figure 3 of the first paper. Notice the lack of right-left symmetry in the heat map of the male's face. I became curious whether handedness could be detected from images. However, no such research seems to exist. This might be a great project for publication or hackathon.

Additional Resources:

1. See Additional Resources listed in tonight's Meetup description: https://www.meetup.com/Paris-Machine-Learning-Study-Group-in-English-Meetup/events/261892701

2. Previous Meetup about the primary paper: https://www.meetup.com/Paris-Machine-Learning-Study-Group-in-English-Meetup/events/243370560

3. Where can I find the AI 'gaydar' implementation that identifies gay people through faces?: https://www.quora.com/Where-can-I-find-the-AI-gaydar-implementation-that-identifies-gay-people-through-faces

4. AI Defines Sexual Orientation. Game of Thrones. Men actors: https://www.youtube.com/watch?v=_UP5c2Oj9mY