GANGU is a toy made with AI technology Generative Adversarial Network (GAN). With GAN, two adversarial neural networks are used to create and refine data: Generator (network that generates​ new​ data) and discriminator (network that classifies data).​ ​GAN’s work process is comparable to a manufacturer of counterfeit bills and the police that investigate the authenticity of the bills. We strongly felt the potential of GAN as an AI technology, and assessed its level of intelligence to that of children aged 4-5. GANGU is a toy created by preschoolers who can draw a car and autonomously recognize the drawing as a car. Imagine children playing by riding toy cars as if they are actually speeding down a highway. Fundamentally, toys are simultaneously fake and real. To adults, a toy car is a fake car. To children, a toy car is a real super car. The definition of real and fake is made by humans, society, rules. What is authentic, what is counterfeit? What is reality, what is virtual? With GANGU, we want to question the essence of what authenticity means.