Skip to main content

Watch and learn: Machines taught to learn through observation

turing learning swarmrobots
University of Sheffield
We take learning for granted. From infancy, humans have an innate ability to learn from our observations, enabling us to mimic conversation, avoid unpleasant environments, and recognize that good behavior begets rewards. For machines, these tasks are hardly as easy.

“Turing learning has outperformed a traditional machine learning method when it comes to accurately predicting what a swarm of robots do.”

Recommended Videos

But researchers from the University of Sheffield say they have developed a method that helps artificial systems learn by observing the world around them. It may lay a new foundation for predictive and learning technologies.

Dubbed “Turing learning,” this new method was inspired by the famous Turing test, which was first proposed by the father of computer science, Alan Turing, in 1950.

In the traditional Turing test, a human interrogator converses with two unseen entities, one being human and the other being an artificial intelligence. If the interrogator cannot distinguish the human from the machine, the machine is said to have passed the Turing test.

In the Turing learning study, the human interrogator was replaced by an algorithm, the human subject was replaced with a swarm of robots, and a computer simulation attempted to replicate the robots’ movement, just as the Turing test AI attempts to replicate human speech. By analyzing the movement patterns of the swarm and the simulation, the algorithm tried to determine which was which. The algorithm was rewarded for guessing correctly. The simulation was rewarded for fooling the algorithm.

“Turing learning may also be used one day to predict the behavior of humans, no matter whether they are in the real world or the Internet.”

Dr. Roderich Gross and his team used Turing learning to train the algorithm and train the simulation, such that the two learned their tasks in conjunction with each other. In a paper published online at arXiv, they describe how the learning method allowed the systems to make more accurate simulations of the robot swarm through the simple act of observation without needing human input.

“Predicting swarms is particularly challenging, as the motion of each individual within the swarm is different, every time you look at it,” Gross tells Digital Trends. However, “Turing learning has outperformed a traditional machine learning method when it comes to accurately predicting what a swarm of robots do.”

The study observed only robot swarms which, although thought provoking, is of limited practical value. Still, the researchers think Turing learning may offer a new method of prediction and learning for behavior in the natural world and even the world wide web.

“We would like to see whether it can predict what natural swarms do — from colonies of bees to schools of fish,” Gross says. “Turing learning may also be used one day to predict the behavior of humans, no matter whether they are in the real world or the internet.”

Dyllan Furness
Former Digital Trends Contributor
Dyllan Furness is a freelance writer from Florida. He covers strange science and emerging tech for Digital Trends, focusing…
This tech was science fiction 20 years ago. Now it’s reality
Hyundai Wearable Exoskeleton, assistive tech

Twenty years really isn’t all that long. A couple of decades ago, kids were reading Harry Potter books, Pixar movies were all the rage, and Microsoft’s Xbox and Sony’s PlayStation were battling it out for video game supremacy. That doesn’t sound all that different from 2021.

But technology has come a long way in that time. Not only is today’s tech far more powerful than it was 20 years ago, but a lot of the gadgets we thought of as science fiction have become part of our lives. Heck, in some cases, this technology has become so ubiquitous that we don’t even think about it as being cutting-edge tech.

Read more
Scientists are using A.I. to create artificial human genetic code
Profile of head on computer chip artificial intelligence.

Since at least 1950, when Alan Turing’s famous “Computing Machinery and Intelligence” paper was first published in the journal Mind, computer scientists interested in artificial intelligence have been fascinated by the notion of coding the mind. The mind, so the theory goes, is substrate independent, meaning that its processing ability does not, by necessity, have to be attached to the wetware of the brain. We could upload minds to computers or, conceivably, build entirely new ones wholly in the world of software.

This is all familiar stuff. While we have yet to build or re-create a mind in software, outside of the lowest-resolution abstractions that are modern neural networks, there are no shortage of computer scientists working on this effort right this moment.

Read more
Inside the rapidly escalating war between deepfakes and deepfake detectors
Facebook Deepfake Challenge

Imagine a twisty-turny movie about a master criminal locked in a war of wits with the world’s greatest detective.

The criminal seeks to pull off a massive confidence trick, using expert sleight of hand and an uncanny ability to disguise himself as virtually anyone on the planet. He’s so good at what he does that he can make people believe they saw things that never actually happened.

Read more