AI ‘hot or not’ tool rates celeb faces: Sydney Sweeney and more shocking results

nypost.com:

HotChat 3000 on Thursday, but its ranking output might need some recalibration. The software allows users to input a photo, whether of themselves or a celebrity and regurgitates a rating from one to 10, matching the user with someone of similar attractiveness – according to them.The robo-Y2K interface – a blinding neon green – pays homage to the world wide web’s era of Hot or Not and Omegle.

MSCHF isn’t exactly “re-inventing the wheel,” as they say, but rather drawing on inspiration and methodology from the past.“HotChat 3000 is built from pre-existing stock parts: It is a snapshot of what a convenient assemblage of publicly available models and datasets determine human hotness to be,” reads the company’s statement provided to The Post.

In other words, artificial intelligence has learned our beauty standards. The company utilized CLIP, Open AI’s tool which is trained with 400 million image-text pairs, according to the siteBut MSCHF went one step further to refine their hotness ratings, utilizing additional datasets SCUT-FBP5500, which predicts facial beauty, and Hot or Not.Because the software is trained by humans’ preconceived beliefs of what is not and what is not, it holds implicit bias due to our cultural stereotypes.“Beauty standards do exist and machine learning models trained on vast amounts of data inevitably learn them,” MSCHF’s statement reads, acknowledging the AI’s “highly imperfect” nature.The chatbot is the latest brainchild of the eccentric brand, which dabbles in a bit of everything, from fashion to technology.

Most known for being the masterminds behind those Big Red Boots, MSCHF also created the 2 Million Dollar Puzzle, where players have a slim chance at winning big if they can piece together the.