In the Media

After observing the behavior of 13 AI agents on a simulated desert island, MIT researchers suggest that allowing agents to explore “without a goal” could be an effective way to teach AI agents new skills, reports Aditya Mehta for TechCrunch. “We normally develop artificial agents by deciding in advance what they should practice, what they should optimize, and what counts as progress,” the researchers write. “Play suggests another regime: give an agent a sufficiently rich and safe world, let it choose what is interesting, and allow useful knowledge to accumulate from the history that follows.”