Artificial intelligence DeepMind no longer needs ‘human input’ to learn
Artificial intelligence DeepMind developed by Google, no longer needs the support of their creators in order to develop and improve further. This has been achieved through the implementation of a new system of Differential Neural Computer (DNC), which combines the ability of computers to store large amounts of information, logic, artificial intelligence, and the ability of a neural network to quickly search for data warehouse fragments.
“This model is able to learn as a neural network, but it is able to store large amounts of data like computers”, — explained the leading developer DeepMind Alexander graves and Greg Wayne.
Like the brain, the neural network uses a network of interconnected nodes in order to stimulate the proper centers required to accomplish the task. Artificial intelligence optimizes the performance of the nodes in order to find the fastest solution in this case will be correct. Over time, the AI will use experience in order to look for the right answer faster and more efficient.
The developers of Google gave a vivid example of the new system. After the DNC explained how to construct a family tree, the artificial intelligence was able to build multiple relationships between all the parties, while optimizing and organizing your memory in the future you could quickly find the necessary information. After that, artificial intelligence showed a map of the London underground, and he began to look for additional paths and the relationships between them, in the event that be able to lay the correct route from one point to another.
Instead of trying to calculate every possible outcome of a situation in search of the right decisions, DeepMind could find the answer based on their previous experience, to detect it in its internal memory and linked to the current situation. At the same time people don’t have to give him hints or to feed new input data in huge quantities. For scientists this is a huge achievement in the development of artificial intelligence. It worked DeepMind, when earlier this year, defeated the world champion in the game. But that was only the beginning. And what will happen next – time will tell.
MTCHT
ICT
TECHNOLOGICAL INNOVATIONS
POST
ABOUT US
NEWS
INTERESTING
INTERVIEW
ANALYSIS
ONLAIN LESSONS
