Home

Sweet Dreams, Robot

Dec. 6, 2016

Google is teaching robots to dream. So what does this new robot behavior mean for humans?

In a recent Beyond The Factory Walls “Shop Talk” episode in which Beth Stackpole and I chat about some of the interesting and sometimes unusual technologies in development, a discussion about applying emotion to artificial intelligence led to a short reference to the show “Westworld” and my fear that we are crossing a line a little bit by giving robots human capabilities. Specifically, in the HBO show, the robot creator (Anthony Hopkins) programs “reveries” into the robot code and now some are “remembering” some of the horrible things that have happened to them. And, well, they’re not happy.

Yes, it’s a show, but it is becoming all too real now that Google is giving its DeepMind neural network the ability to dream. The goal is to enhance the rate of learning for artificial intelligence. According to an article in Futurism, the dreams are based on scenes from video games to help AI learn to recognize challenges in the game and help it make better decisions. And the dream training has increased its performance, the article said.

It’s all very experimental at the moment, but, honestly, whether it’s a Socially Aware Robot Assistant (S.A.R.A.) or a dreaming robot, the lines between man and machine are beginning to blur.

About the Author

Stephanie Neil | Editor-in-Chief, OEM Magazine

Stephanie Neil has been reporting on business and technology for over 25 years and was named Editor-in-Chief of OEM magazine in 2018. She began her journalism career as a beat reporter for eWeek, a technology newspaper, later joining Managing Automation, a monthly B2B manufacturing magazine, as senior editor. During that time, Neil was also a correspondent for The Boston Globe, covering local news. She joined PMMI Media Group in 2015 as a senior editor for Automation World and continues to write for both AW and OEM, covering manufacturing news, technology trends, and workforce issues.