DeepMind, the AI research firm that is owned by Google’s parent company Alphabet, has created a computer that can learn and accomplish tasks from its own memory. The computer uses a neural network that can store and recall facts. This allows it to, for example, find the best route from one point to another on London’s famed Underground subway without any prior programming.
DeepMind calls its newest project the differentiable neural computer. Here’s a very quick description of how it works:
When we designed DNCs, we wanted machines that could learn to form and navigate complex data structures on their own. At the heart of a DNC is a neural network called a controller, which is analogous to the processor in a computer. A controller is responsible for taking input in, reading from and writing to memory, and producing output that can be interpreted as an answer. The memory is a set of locations that can each store a vector of information.
Over time this controller can come up with better answers to questions by using its memory. One experiment used the map layout of the London Underground as its basis:
When we described the stations and lines of the London Underground, we could ask a DNC to answer questions like, “Starting at Bond street, and taking the Central line in a direction one stop, the Circle line in a direction for four stops, and the Jubilee line in a direction for two stops, at what stop do you wind up?” Or, the DNC could plan routes given questions like “How do you get from Moorgate to Piccadilly Circus?”
DeepMind’s new computer can also figure out relationships in a family tree with the same learning and memory method, along with solving puzzles. The AI firm hopes that development of DNCs will only create better computers but might also help in finding out how memory works in humans as well.