A robot with a broken leg taught itself to walk again thanks to a brain powered by artificial intelligence (AI) that never stops learning.
The bot has a brain that is constantly evolving and adapting to new circumstances.
A a new study has found mathematical formulas that allow AI brains to keep learning.
The ground-breaking results mean that the robots are able to adapt to their surroundings.
Most AI rely on neural networks, which are algorithms based on brains like a humans.
The main difference is that a human continually learns and adapts its thinking and behaviour, while AI brains can’t.
But to test this, researchers of the new study created nets with "Hebbian rules" — which are formulas that enable AI to continually learn.
In an experiment, researchers had two robots – one with the Hebbian network, and one with the usual dynamical weight, reports Science Mag.
For both bots scientists removed a leg, leading to two very different results.
AI 'dog' to explore Mars' buried caves and tunnels in latest Red Planet breakthrough
The advanced creature was able to quickly adapt thanks to a constantly learning brain and quickly figure out how to walk.
For the secondary robot, it was unable to do so and flailed around on its back.
After initially struggling, the Hebbian bot was able to walk nearly seven times as far, researchers said.
The findings, presented to the Conference on Neural Information Processing Systems, could improve algorithms in future AI robots, leading to more advanced creations.
Future AI could benefit from such Hebbian networks, which would be useful for advancing robotics.
AI could use the system to learn new languages, driving skills, or recognise images.
Source: Read Full Article