By combining complex computer data processing with artificial neural networks, a group of artificial intelligence researchers in London just created what could be the Jason Bourne of neural computers.
Their smart hybrid learning machine can understand family trees, solve complicated puzzles and determine the best route on the London Underground - without having seen the transportation system before.
The machine's development was led by a team from the artificial intelligence research company DeepMind that included research scientists Alex Graves and Greg Wayne as well as co-founder and CEO Demis Hassabis.
When it comes to fake brains, computers can do complex data processing but they need to be manually programmed. Artificial neural networks don't have that drawback, but they also don't have the ability to allocate new storage dynamically when memory demands increase for a task.
RELATED: Mind Meld! Top Brain-Controlled Techs
Recently the DeepMind team figured out how to unite the two by building a differentiable neural computer or DNC for short. They essentially made a new artificial neural network and gave it an external memory structure. With this sophisticated configuration, the system doesn't need any programming, just a little guidance.
"When trained with supervised learning, we demonstrate that a DNC can successfully answer synthetic questions designed to emulate reasoning and inference problems in natural language," the DeepMind researchers wrote in a Nature journal article published this week (abstract). In other words, their machine picks stuff up in a human way.
WATCH VIDEO: Why You Shouldn't Fear Artificial Intelligence