Robots don’t know how to cope with damage. Unlike humans, machines are unable to think of creative solutions beyond their basic functions. When you sprain your finger from playing video games all night, your other capable appendages compensate for the impaired movement. But when droids encounter a similar situation, they will continue trying to execute the same action because that’s all they know how to do (poor bots).
To help machines survive like animals, researchers from Pierre and Marie Curie University and the University of Wyoming developed a new algorithm that allows robots to adapt to injuries. The breakthrough brings machines one step closer to having Terminator-like capabilities. Skynet, though, is still far from reality.
“When injured, animals do not start learning from scratch,” said senior author Jean-Baptiste Mouret. “Instead, they have intuitions about different ways to behave. These intuitions allow them to intelligently select a few, different behaviors to try out and, after these tests, they choose one that works in spite of the injury.”
This video shows the trial-and-error algorithm in action:
Based on the footage, you’re able to see the thought process that the bot applied to reach its goal of walking straight. It creates a map of options and tests each one without hesitation. Because bots can’t feel pain, they’re able to work through the solutions in a methodical manner. Surprisingly, the same program can be applied to help machines learn about fresh environments and situations. They can even create new behaviors using the algorithm!
An obvious drawback is the need for the machine to experiment with each option on the map, instead of “thinking” about the solutions. Going back to the case of the sprained finger, it is unnatural for a human to forcefully keep trying to use the impaired body part, knowing that there’s something wrong with it. Doing so would cause more injury. But at the rate that scientists are developing the algorithm, robots may soon be able to repair themselves and carry on like nothing happened.
“We performed experiments that show that the most important component of Intelligent Trial and Error is creating and harnessing the prior knowledge contained in the map,” highlighted Jeff Clune, Assistant Professor from the University of Wyoming, who is also a participant in the study. “It could enable the creation of robots that can help rescuers without requiring their continuous attention. It also makes easier the creation of personal robotic assistants that can continue to be helpful even when a part is broken.”
Dark Robot Intuition
The advancement is very impressive. Manufacturers can drive down costs and boost the shelf life of gadgets by incorporating the technology in everyday tools. It’s just that the idea of a rogue android that doesn’t die (even after taking numerous hits) doesn’t sit well with most tech enthusiasts who understand the possibility of A.I. taking over the world. With some people empathizing with soul-less bots, it’s clear that the complexity of our relationships with machines has already evolved to a new level of reality.