Although robots are not capable of imagining themselves in the same way as the humans, however, this would be a reality soon. And this ability of machines would lead to their self-awareness.
Recently a team of engineers from Columbia University has found success in developing a robot which is capable of building its own image from scratch. This is a big achievement considering this fact that today’s robots are yet to learn the process of creating their own images. Before this, the robots used to rely on the trial and error method which is pretty time consuming or the models installed by humans sometimes.
These newly built robots first do the intense computing for one and a half day. Then on the basis of these calculations, these machines learn about their identity without any prior knowledge of Science concepts. They use their own technique of self-simulation so that they could succeed in adapting to a variety of different environmental changes around them.
Hod Lipson, a mechanical engineering professor at Columbia University who was also a part of this team of engineers explained that it is important that robots become independent and capable of adapting to different conditions nowadays. This new robot was designed and developed at Lipson’s lab named Creative Machines Lab.
While building this robot, the team first focused on designing a robotic arm which had freedom of movement of four degrees. Then it started making some random moves and logged 1000 trajectories successfully. Each trajectory consisted of positional data points numbering to 100. Through this data, the robot managed to build its self-model.
Although the first self-image was not that accurate to enable the robots to identify itself directly and make it understand how these dots were connected to each other. Then, the team initiated an algorithm of deep learning, through which the robot carried on to predict its self-model for next 35 hours. Finally, it produced a self-image which was one and a half inch long.