Rather than a neural network, the team used a so-called Bayesian program learning framework. Because the algorithm is based on probability and guessing, it's using a cognitive process like humans and not a typical rote computer method.
When the machine "drew" characters on the screen, each one was slightly different but still identifiable like the ones we would draw. As a result, only 25 percent of judges who compared the samples to human-drawn characters were able to tell the difference. Author Joshua B. Tenenbaum told the NYT that "it's amazing what you can do with lots of data and faster computers. But when you look at children, it's amazing what they can learn from very little data. Some comes from prior knowledge and some is built into our brain."