DeepMind a company that was acquired a few years ago by Google, is back in the news thanks not only to the enormous capabilities that its artificial intelligence system is showing, which is already capable of beating the best humans in different games, but also because of its new progress where, today, it seems, the platform is able to learn and acquire new knowledge on her own, without human intervention.
In order to take this new step, apparently and as discussed on the official project page, DeepMind researchers decided to test and create a differentiable neural computer, better known by its acronym in English DNC, a system that basically stands out for being able to solve any type of problem without having prior knowledge of them. Going into a little more detail, what differentiates the DNC is that it is capable of storing data like a conventional computer and combining it with a neural network computer system.
The DeepMind platform is already capable of acquiring knowledge without human intervention of any kind.
Like a human brain, this neural network uses a series of node connections to stimulate the specific centers needed to complete a task. In this case, artificial intelligence is nothing more than the optimization of nodes in order to find a faster solution to the problem in question. Thanks to this advance, DeepMind is now able to solve situations such as planning the best route between London Underground stations or directly elaborating relationships between relatives in family trees.
As you have commented so much Alexander Graves as Greg wayne, DeepMind researchers:
When we designed DNC, we wanted machines capable of learning to navigate and form complex data structures on their own. At the heart of a DNC is a neural network called a controller, which is responsible for taking input, reading and writing memory, and producing what can be interpreted as a response.
Further information: DeepMind