Robot Uses AI to Learn 1,000 Tasks in a Single Day

­AI is rapidly expediting the machine-learning process, and to that end, a new development described in Science.org is both illuminating and slightly terrifying – researchers managed to teach a robot 1,000 physical tasks in a single day, each from a single demonstration.

TechRadar points out that, prior to the onset of AI, machine learning was slow and laborious, with umpteen demos resulting in a single, highly-specific area of expertise – hence their deployment in factories and assembly lines.

AI has begun to change that painfully arduous process for the better.

Case in point – researchers developed something called Multi-Task Trajectory Transfer (MT3), an imitation learning method which, according to an article in Science.org, is “based on decomposition and retrieval.”

“MT3 learns everyday manipulation tasks from as little as a single demonstration each while also generalizing to previously unseen object instances. This efficiency enabled us to teach a robot 1000 distinct everyday tasks in under 24 hours of human demonstrator time,” they claim.

Essentially, the robot breaks each action into smaller phases, and the result is something closer to how we learn things vs, well, robots. Obviously, an AI-enabled robot that can learn as fast (or faster) than us and think faster than us would be exponentially more valuable to both industry and consumers.

It also potentially edges closer and closer to an intelligence that can perform all human functions faster and better than us, but we’re not there yet.