Picture of the GANPaint tool where users can paint various features, such as trees, grass, or doors, onto a picture of a church.


Generative Adversarial Networks (GANs) are one of the hottest topics in AI. A pair of neural networks work together to create output so realistic that it can be difficult to distinguish from content produced by humans. While GANs can generate text, audio, and other forms of data, they initially became renowned for their applications in image and video. In a project with the Quest Bridge, undergraduates are helping to create an extension to the popular coding platform for kids, Scratch, that teaches about GANs through creative play. For example, one of the features in this extension incorporates GANpaint, an application developed in the lab of Antonio Torralba in CSAIL by David Bau and collaborators, which encourages users to develop an intuition for how GANs work by playing with GAN-generated images. Users can then turn the images they’ve made with GANpaint into a background for their Scratch projects.



Katherine Gallagher

MIT Quest for Intelligence

Brent Samuels


Jamison Rich


Maya Nigrin


Philip Tegmark