A Robot Data Startup Raises $60 Million

Companies developing AI models to power humanoid and other robots have been hard at work collecting videos and other data for training their models, even paying people to record themselves completing tasks in homes and workplaces. 

As these data-collection efforts start to pay off, and robots take on some of those tasks, the robot developers could start prioritizing quality of data over quantity. That’s a scenario that Encord, a five-year-old startup, is already preparing for with its software for helping AI developers manage data.

According to Encord’s co-founder and co-CEO, Ulrik Hansen, companies developing AI, including for robots, increasingly want to weed out low-quality data. They’re rejecting videos in which a person’s hands are out-of-frame or a person is watching TV rather than doing tasks an AI model could learn from, for instance, he said.