Meet BRETT, the robot that learns by trial and error

Robots are remarkably efficient at handling pre-programmed tasks ranging from remarkably simple to incredibly complex. Tasks such as vacuuming a floor, flying an airplane, or near-complete assembly of modern automobiles are all currently within the realm of robot capabilities.

The one inefficiency in modern robots is the reliance on humans to provide them with the correct sequence of 1’s and 0’s needed to carry out a task.

This could soon change.

Meet BRETT.

BRETT – or Berkeley Robot for the Elimination of Tedious Tasks – is a creation of the Berkeley Robot Learning Lab that uses deep learning techniques in order to complete seemingly simple tasks without any direct input from humans. Tasks like connecting two LEGO pieces, inserting a peg into a hole and unscrewing a water bottle might seem simple to your average toddler, but provide a challenging puzzle for artificial intelligence and machine learning algorithms.

When pre-programmed, these tasks are easily executed by a robot, but BRETT differs in that the machine learns how to execute the steps necessary for solving these simple problems on his own. In a few years, who knows what BRETT, and others like it, will be capable of.