This decade’s data is what led to the University’s new experiment in artificial intelligence.

Dr. Finn and his team created a neural network, a mathematical system that could learn skills from a huge amount of data. Through examples in thousands of cat photos, the neural network can learn to recognize cats. By analyzing hundreds of old phone calls, he can learn to recognize spoken words. Or, by teaching teaching assistants how to evaluate coding tests, they can learn to evaluate these tests themselves.

The Stanford system, learning from a decade of possibilities, spent hours analyzing examples of old Midwest. Then he was ready to learn more. When given only a few additional examples of the new exam to be given this spring, he can quickly learn the task at hand.

“He sees a variety of problems,” said Mike Wu, another researcher working on the project. “It can then be adapted to problems it has never seen.”

This spring, in you, the system provided 16,000 pieces of feedback, and according to a study by Stanford researchers, students agreed with 97.9 percent of the response over time. By comparison, students agreed with human instructors ’response to 96.7 percent of the time.

Mr. Fam, an engineering student at the University of Lund in Sweden, was amazed that TechnoSoGA did such a good job. However, the automated tool was unable to evaluate one of its programs (probably because it wrote a snippet of code unlike anything seen by AI), both of which identified specific errors in its code, including what is known in computer programming and mathematics as a fence. Post as an error, and suggest ways to correct them. “It’s rare that you get such a well-thought-out response,” Mr. Fame said.

The technique was effective because its role was very sharply defined. While taking the test, Mr. Fam wrote the code with very specific objectives, and there were many ways that he and other students could go wrong.

But given the true data, neural networks can learn many functions. This is the same basic technology that recognizes faces in photos you post on Facebook, recognizes barking commands on your iPhone, and translates from one language to another on services like Skype and Google Translation. The hope for the Stanford team and other researchers is that these techniques can automate learning in many other ways.