Advertisement

Google's memory-boosted AI could help you navigate the subway

It's better at handling complex tasks than conventional neural networks.

REUTERS/Neil Hall

Modern neural networks are good at making quick, reactive decisions and recognizing patterns, but they're not very skilled at the careful, deliberate thought that you need for complex choices. Google's DeepMind team may have licked that problem, however. Its researchers have developed a memory-boosted neural network (a "differentiable neural computer") that can create and work with sophisticated data structures. If it has a map of the London Underground, for example, it could figure out the quickest path from stop to stop or tell you where you'd end up after following a route sequence.

The key is how the AI uses its memory. The computer's controller is figuring out how to use memory as it goes along -- it's learning how to get ever closer to the correct answer without being explicitly told how to get there. A typical neural network wouldn't even have that memory to work with, so it either wouldn't hold the information or find a general way of reasoning that translates to different circumstances.

This wouldn't just be useful for navigation. You could ask the computer to identify relatives in a family tree using only basic knowledge about the relationships, or solve intricate puzzles with varying goals. All told, the very potential of neural networks is about to expand. You could see AI applied in situations where it was previously impossible.