The data from this decade led to the university’s new experiment in artificial intelligence.
Dr. Finn and his team created a neural network, a mathematical system that can learn skills from large amounts of data. By pinpointing patterns in thousands of cat photos, a neural network can learn to identify a single cat. It can learn to recognize spoken words by analyzing hundreds of old phone calls. Or, by examining how teaching assistants evaluate coding tests, he or she can learn to evaluate these tests on their own.
The Stanford system spent hours analyzing examples from the old midterms, learning from a decade’s worth of possibilities. Then it was ready to learn more. When given a few additional examples from the new test offered this spring, it could get the job done quickly.
“It looks at a variety of problems,” said Mike Wu, another researcher who worked on the project. “Then it can adapt to problems it has never seen before.”
According to a study by Stanford researchers, this spring, the system provided 16,000 pieces of feedback, and students agreed with 97.9 percent of the feedback. By comparison, students agreed with human instructors’ feedback 96.7 percent of the time.
Mr Pham, an engineering student at Sweden’s Lund University, was surprised that the technology was working so well. Although the automated tool was unable to evaluate one of his programs (probably because he had written a snippet of code unlike anything the AI had seen), it identified specific bugs in his code, including those in computer programming and mathematics. what is known. A fence post error, and suggested ways to fix them. “It’s seldom you get such a well thought out response,” said Mr. Pham.
The technology was effective because its role was so sharply defined. In taking the test, Mr. Pham wrote code with very specific objectives, and there were only so many ways he and the other students could be wrong.
But given the right data, neural networks can learn a wide variety of tasks. It’s the same fundamental technology that recognizes faces in photos you post on Facebook, recognizes commands you come to your iPhone, and translates from language to language on services like Skype and Google Translate. . For the Stanford team and other researchers, the hope is that these technologies could automate learning in many other ways.