Internet giant acquired a couple of years ago DeepMind, the company that owns a technology aimed at making computers think like humans achieve, which is specially applied to the field of video games. In recent years, this technology has made incredible fruits, as the victory against the world champion Go to AlphaGo hands.
Well, it seems that researchers created a differentiable neuronal DeepMind computer, able to solve problems without having prior knowledge of them. In other words, the DNC uses the storage capacity of conventional computers and computer system combined with a neural network.
Like the brain, the neural network uses a series of nodes interconnected by stimulating specific centers needed to complete a task. In this case, the artificial intelligence is the optimization of the nodes to find the fastest solution to deliver the desired result.
With this development, the DeepMind platform is now capable of solving situations like planning the best route between London Underground stations, or develop relationships among relatives in family trees. Even the system is able to solve a block puzzle game using reinforcement learning.
You may also like to read another article on Web2GB: Welcome, speed a new standard five times the speed of Gigabit Ethernet
“When we designed DNC, we wanted machines that can learn to sail and form complex data structures on their own. In the heart of a DNC a call neural network controller, which is responsible for taking input, read and write memory, and produce what can be interpreted as a response”, said Wayne Alexander and Greg Graves, researchers DeepMind.
Thus, instead of having to learn every possible outcome to find a solution, DeepMind can find an answer based on previous experience, unearthing what you need to know your own internal memory. You can even say it was the same process that DeepMind could beat a human champion ‘Go’, a game with millions of movements and an infinite number of combinations.