25/01/2023
Robots struggle with simple
Scientists and engineers at the UC Berkeley Robotics School designed the Berkeley Robot for the Elimination of Tedious Tasks (Brett). Brett’s job was to pick up and fold a towel. After years of effort, Brett’s best time was 20 minutes. More recent improvements, involving deep learning algorithms, have achieved best times of 1.5 minutes. However, Brett still gets stumped routinely when the laundry is messy. Overall, humans of all ages and skills outperform Brett at folding towels in a matter of seconds.
Brett’s towel tribulations fall within the space of Moravec’s paradox: we can design algorithms that solve complex problems like winning at world-class chess or flying an aircraft but are unable to complete simple tasks like tying shoelaces or folding a crumpled towel. Simple tasks that involve irregularity, complexity, creativity, ethical judgements and emotional intelligence are often beyond the reach of robotics.
However, there is still much scope for artificial intelligence to automate previously human-performed tasks soon. Artificial intelligence has amazed us in recent history by solving complex problems that were once considered impossible for a robot.
Find more via the link in our bio.