Microsoft has teamed up with MIT might to develop a model that uses artificial intelligence to catch virtual “blind spots”. Basically, the AI compares a human’s actions in a given situation to what it would have done, and modifies its behavior based on how closely it matches the response. Put simply, if a self-driving car doesn’t know how to pull over in an emergency situation, it could learn how to do so by simply observing a human driver moving to the side of the road. However, if the AI is wrong, a human driver can also step in to correct it. “The model helps autonomous systems better know what they don’t know. Many times, when these systems are deployed, their trained simulations don’t match the real-world setting [and] they could make mistakes, such as getting into accidents. The idea is to use humans to bridge that gap between simulation and the real world, in a safe way, so we can reduce some of those errors,” said first author Ramya Ramakrishnan from Computer Science and Artificial Intelligence Laboratory at MIT. Read more for a video and additional information.
“Researchers even have a way to prevent the driverless vehicle from becoming overconfident and marking all instances of a given response as safe. A machine learning algorithm not only identifies acceptable and unacceptable responses, but uses probability calculations to spot patterns and determine whether something is truly safe or still leaves the potential for problems. Even if an action is right 90 percent of the time, it might still see a weakness that it needs to address,” reports Engadget.