Scientists Create a Self-Aware Robot That Can Repair Itself

Scientists at Columbia University developed a self-aware robot that became entirely autonomous in just 35 hours. It not only learned what it was, it also learned how to repair itself.

For decades, we always imagined a self-aware robot in the same way we do aliens and spaceships, as characters from sci-fi movies. So, to catch a glimpse of both, we usually have to see one Will Smith movie or another.

Well, that’s about to change.

According to a publication in Columbia Engineering, engineers at Columbia University have created a self-aware robot that operates on its own and is capable of repairing itself.

The engineers first built a mechanical arm with no prior knowledge of physics, motor dynamics, or geometry. Initially, it was just that, a machine that could not tell that it was an arm. But after 35 hours, the robot arm figured out its purpose using deep learning.

Read More: Why Do People Bully Robots?

The robot created a self-simulation which it used to contemplate and adapt to situations. As a result, not only could it handle new tasks, but it could also detect and repair damages in its body.

It started with the basic task of collecting objects from specific locations and moving them into a receptacle. The robot arm did this with a 100 percent success rate.

Then it got better. The robot was able to say “Hi” by writing with a marker. Finally, the researchers decided to test if the robot could detect damages to itself.

They printed a 3D part that’s slightly deformed and attached it to the robot. Within a short while, the robot detected the change and updated its self-model to account for the changes. After that, it continued with the pick-and-place task with little decline in performance.

“This is perhaps what a newborn child does in its crib, as it learns what it is,” said Professor Hod Lipson, Head of the Creative Machine Labs at Columbia.

Today’s robots operate by depending on a human model. But Hod Lipson thinks robots will become independent and adapt faster when they learn to simulate themselves.

The Columbia University professor also believes that robotics and AI may provide a better understanding of the age-old puzzle of consciousness.

Now the researchers are exploring whether robots can model not just their bodies, but mind too. In other words, they are trying to create machines that can think and are aware that they are feeling.