Ability to imagine and recognize oneself is only attributed to humans. However, things are about to change as a new ‘self-aware’ robot has been created that ‘imagines’ itself.
Researchers from Columbia University have created a new ‘reflective robot’ that uses the ability of self-awareness to adapt to various situations and tasks. The robotic arm had no prior knowledge or experience before imagining itself.
The robot also had no knowledge of geometry, motor dynamics or physics, and did not know anything about its shape or dimension. Within a day of intense computing, called ‘babbling’, the robot created a self-simulation and used it to help itself take on different assigned tasks.
The robotic arm used its self-awareness trait to help itself detect and repair damage to its own body. For it to reach and create its own self-simulation, the arm used a combination of deep learning and machine learning techniques.
For the study, the team made use of a four-degree-of-freedom articulated robotic arm. At first the robot moved randomly and inaccurately, but over a period of 35 hours, the bot learned through these movements, ultimately making its own discovery.
Though the robotic arm is not near the ability of humans and other animals, researchers say that it could be similar to how human infants learn early motor skills.
“While our robot’s ability to imagine itself is still crude compared to humans, we believe that this ability is on the path to machine self-awareness,” said Hod Lipson.
Self-awareness will lead to more resilient and adaptive systems but also implies some loss of control. It’s a powerful technology, but it should be handled with care,” Lipson added.