Map service: AI evaluates the attractiveness of guests
A programmer has analyzed the profile pictures of people who have rated locations on Google Maps and created a map from them.
(Image: Screenshot/LooksMapping)
Thanks to AI, there is now an attractiveness map for Manhattan's restaurants—LooksMapping. You can also see the average age and gender of visitors. It's not as if you're a potbellied barbecue plate eater and you run into professional athletes with protein shakes in a bar or a hip tourist seeking food spots for Instagram and end up at the grannies in the café.
To create the map, programmer Riley Walz analyzed the profile pictures of people who have rated a place on Google Maps. Red places are where you'll find particularly hot people, while blue is less attractive. For Manhattan, for example, this means that the entire Financial District at the northern tip of the borough attracts less attractive people than the trendy East Village. Neither seems surprising.
Videos by heise
It gets really deep blue towards Harlem. And therein lies the crux of the map. Many dark-skinned people and those with a Hispanic background live in Harlem. The AI's assessment indicates a bias—you could also say it is racist. The creator writes directly on the website where the map is embedded: The model is certainly biased and flawed. “But we rate places according to the people who are there. We've always done that. And don't we also have mistakes?” The website only underpins everyday, superficial calculations with figures, writes Walz. “It holds up a mirror to our collective vanity.”
Clip evaluates faces instead of captions
Walz has recorded how he proceeded for the rating card in a published paper. He analyzed almost 600,000 accounts with profile pictures of Google Maps users. Each image was evaluated using an AI model developed by OpenAI. Clip stands for Contrastive Language-Image Pre-Training and is actually designed to select the most appropriate caption for an image. For LooksMapping, Clip had to match sentences such as “She is attractive and beautiful” or “She is unattractive and ugly” as well as “A young person” or “An old person” with the images. Only after 50 faces per restaurant had been evaluated was the average calculated and an indication given on the map.
In addition to Manhattan, there are also maps for Los Angeles and San Francisco. While at least downtown Manhattan is rather red overall, there are fewer attractive places in San Francisco as a whole. Surprisingly, it seems that particularly few attractive people go to restaurants in Los Angeles.
(emw)
