Computer Learns How to Mimic Human Brain

Researchers at Los Alamos National Laboratory and the New Mexico Consortium are working to create computers that can act much like the human brain.

The human brain is an incredible computer. It requires little energy to run and can sort through incredible amounts of data quickly enough for us to react on. By an early age humans can easily tell the difference between two objects or can tell a moving car from a static background. These kinds of tasks are still challenges for computers to accomplish.

Garrett Kenyon and other researchers in the field of neuromimetic computing have been simulating the brain's neural networks on the Trinity supercomputer at Los Alamos, the fastest computer in the United States. The goal is for the computer to be able to mimic the human brain, enabling machines to learn about their surroundings, interpret data and make predictions the way humans do.

How are they doing this? Computer scientists and neuroscientists at Los Alamos have created a “sparse prediction machine” which is designed to work much like a brain. The researchers expose it to thousands of video clips of moving objects and the sparse prediction machine learns about the visual world simply by watching the videos, much like a child learns by watching the world. Thus, a sparse prediction machine is more biologically realistic in that it can be trained in a similar way to how we learn as humans. 

Using the Trinity supercomputer, the sparse prediction machine eventually was able to watch a video clip and be able to predict the next video frame that will naturally follow. In one example, it was able to continue the motion of a car against a static background. In other words, a computer has been trained to be able to predict what it will see next, which is a great achievement in the field of neuromimetic computing.

Watch the video clip above or read the full article below to learn more!

Full Albuquerque Journal Article: Computer Learns How to Imagine the Future

Los Alamos Press Highlights: Computer Learns How to Imagine the Future

© 2018 New Mexico Consortium