Can AI Detect Sexual Orientation? | Artificial Intelligence Gaydar

Рет қаралды 1,965

3 жыл бұрын

Hello! In this video, I discuss an article critiquing a machine learning model that was designed to detect sexual orientation based on facial images.
Both the model and the critique focus on comparing same-sex and opposite-sex attracted men and women, though there are obviously more than 4 categories of genders/sexual orientations.
Let me know what you think!
Relevant articles:
psyarxiv.com/hv28a/ (original article)
/ do-algorithms-reveal-s... (critique article)
/ physiognomys-new-clothes
blogs.getty.edu/iris/physiogn...
www.wired.com/story/algorithm...
www.technologyreview.com/2019...
www.theverge.com/2020/6/24/21...
LinkedIn: / janice-pan-8baa4910b
Website: www.janpancake.com/
Instagram: / janpancake
xo janpancake

Пікірлер: 13

@AnastasiInTech 3 жыл бұрын

Very interesting video! I think AI would never find out I am a HW engineer :)

@janpancake 3 жыл бұрын

Probably correct, because unfortunately, AI is also very gender biased. If I type ‘the engineer said’ on my iPhone, the suggested words are ‘he’, ‘they’, and ‘I’ 😒

@owen246 3 жыл бұрын

This channel is great! I'm going into a Physics PhD program with a strong crossover between physics and ML. I was wondering if you'd be able to make a video on investing/getting ahead on your finances while on a grad school stipend? Thanks!

@janpancake 3 жыл бұрын

One of my favorite topics! Yes, would love to make a video on this ☺️

@PriDrummond 3 жыл бұрын

Amazing analysis. thanks

@raniermoura3825 2 жыл бұрын

Awesome video!

@2020onwards Жыл бұрын

Informative. Well done. Surprised that this video did not gain greater traction in terms of views.

@Jay-ex9fb 3 жыл бұрын

Wow AI like this can be very bias and can be unethical in some circumstances if we continue to use AI for purposes as these.

@janpancake 3 жыл бұрын

Exactly right!

@simonincognito6286 Жыл бұрын

Nice video, very interesting. However, the critique of the methods or possible reasons for the observed differences does not negate the main conclusions of the original paper: "Given a single facial image, a classifier could correctly distinguish between gay and heterosexual men in 81% of cases, and in 71% of cases for women. Human judges achieved much lower accuracy: 61% for men and 54% for women. The accuracy of the algorithm increased to 91% and 83%, respectively, given five facial images per person."