As good as they've become, artificial intelligence agents are still largely only as good as the data upon which they were trained. They don't know what they don't know. In the real world, people faced with unfamiliar situations and surroundings adapt by watching what others around them are doing and by asking questions. When in Rome, as they say. Experts in educational psychology call this "socially situated learning."
Until now, AI agents have lacked this ability to learn on the fly, but researchers at Stanford University recently announced that they have developed artificially intelligent agents with the ability to seek out new knowledge by asking people questions.
"We built an agent that looks at photos and learns to ask natural language questions about them to expand its knowledge beyond the datasets it was originally trained on," says Ranjay Krishna, first author of a recent study appearing in the journal Proceedings of the National Academy of Sciences. Krishna earned his doctorate at Stanford and is now on the faculty at the University of Washington.
To read more, click here.