DeepMind has released StreetLearn, an interactive environment and dataset consisting of 114,000 panoramic images from Google Street View, to advance the development of AI systems that can navigate using visual cues instead of maps. The dataset includes images from both New York City and Pittsburgh, and DeepMind has created several tasks to test the ability of AI agents, including a courier task that challenges AI agents to travel to a randomly chosen location.
Training AI to Navigate Without a Map
Michael McLaughlin is a research analyst at the Center for Data Innovation. He researches and writes about a variety of issues related to information technology and Internet policy, including digital platforms, e-government, and artificial intelligence. Michael graduated from Wake Forest University, where he majored in Communication with Minors in Politics and International Affairs and Journalism. He received his Master’s in Communication at Stanford University, specializing in Data Journalism.
View all posts by Michael McLaughlin