An artificial intelligence system that reorganizes itself in a similar way to the human brain, evolves and develops new abilities to overcome externally imposed constraints, in a similar way to the human brain during evolution. The new system created by the research group led by Jascha Achterberg and Danyal Akarca from the University of Cambridge and described in Nature Machine Intelligence is an inspiration for future AI systems and will allow us to better understand how it works our brain.
The main objective of artificial intelligence systems is to replicate some characteristics of human intelligence, but there are many differences between the digital world and the natural world, one of which is the ability to change over time.
The biological brain is not only able to solve complex problems, but it does so using very little energy, an ability that is also possible because it is able to rearrange the connections between neurons.
To try to understand how this happens, researchers have developed a type of neural network capable of autonomously transforming in response to what is asked of it and reducing energy consumption as much as possible. A very different target than the usual one, for which artificial neural networks usually have large amounts of energy available.
Using these new constraints, US researchers have shown that if networks are asked to solve difficult problems, for example finding the fastest way out of a maze, while at the same time reducing available energy, they evolve into unexpected ways. Network nodes, as if they were neurons, tend to rearrange connections and learn to manage a larger number of operations than they normally do. A discovery that could improve the design of new artificial intelligence systems, but which, the researchers point out, also opens up many interesting ideas for understanding our brain and why it is structured in the shapes we see. (A2 Televizion)
A2 CNN Livestream
MUND TE LEXONI GJITHASHTU
