A.I. Toys

How does society construct gender through toys? The toys described on these screens were created by a machine learning algorithm trained on real consumer products sold online that have been labeled as “girls” and “boys” toys. After learning from the existing dataset of toys, the A.I. algorithm was asked to invent new toys based on what it learned. The resulting toys have striking differences; girls toys are largely centered on jewelry, domestic chores, dolls and animals, while boys toys are largely centered on weaponry, electronics, cars, and construction. In collaborating with the machine learning model, the resulting toys— derived directly from the toys we produce for children in real life—serve as a reflection of the gendered societal values we place on children through objects of play.

These toy titles and descriptions are generated by a machine learning model trained on real toys currently for sale on Amazon and Target. The top row consists of toys generated from toys labeled “girls”, and the second row are toys generated from toys labeled “boys”. The contrast in content, use of adjectives, and color scheme reflects the way we indoctrinate gender norms through design objects such as toys.

Image credit: Brad Farwell

Prototype of the algorithmically generated toys as a 3d print.

A website with some of the toys generated by the machine learning model: http://ai-toys.net/

Photo credit: Brad Farwell and Ani Liu.

Video by Brad Farwell. As shown at Cuchifritos Gallery + Project Space