AI system mimics human short term memory

Google’s DeepMind has unveiled a prototype computer that attempts to mimic properties of the human brain’s short-term memory. It is a neural network that works with an external memory, resulting in a computer that learns as it stores memories and can later retrieve them to perform logical tasks beyond those it has been trained to do.

A traditional computer neural network consists of interconnected processors that can change the strength of their connection based on external input. This models the plasticity and learning ability of a brain. DeepMind has added a new component based on Turing’s model of computation, in which memory acts as a tickertape that can pass back and forth through a computer, sorting variables for later processing. The component allows DeepMind’s  “Neural Turing Machine” to understand new data as chunks.  The external memory is used to keep the chunks active so it can use them at different points in a calculation.


Posted

in

,

by

Tags: