A quantum neural network can see optical illusions like humans do. Could this be the future of AI?

At first glance, optical illusions, quantum mechanics, and neural networks may seem like unrelated topics. However, in a new study published in APL Machine LearningI used a phenomenon called “quantum tunneling” to design a neural network that can “see” optical illusions in the same way that humans do.

My neural network did a good job of simulating human perception of the famous Necker Cube and Rubin’s Vase illusions, and in fact better than some much larger conventional neural networks used in computer vision.

The work could also shed light on whether artificial intelligence (AI) systems can one day truly achieve something resembling human cognition.

Why optical illusions?

Optical illusions trick our brains into seeing things that may or may not be real. We don’t fully understand how optical illusions work, but studying them can teach us more about how our brains work and how they sometimes malfunction, in conditions like dementia and during long space flights.

Researchers using AI to mimic and study human vision have found that optical illusions are a problem. While computer vision systems can recognize complex objects like paintings, they often can’t understand optical illusions. (The latest models appear to recognize at least some types of illusions, but these findings require further research.)

My research addresses this problem using quantum physics.

How does my neural network work?

When a human brain processes information, it decides which data is useful and which is not. A neural network mimics the brain’s functioning by using multiple layers of artificial neurons that allow it to store and classify data as useful or not.

Neurons are activated by signals from their neighbors. Imagine that each neuron has to climb a brick wall to be activated, and that signals from its neighbors propel it higher and higher, until it finally climbs over the wall and reaches the activation point on the other side.

In quantum mechanics, tiny objects like electrons can sometimes pass through seemingly impenetrable barriers through an effect called “quantum tunneling.” In my neural network, quantum tunneling sometimes allows neurons to pass right through the brick wall to the activation point and fire even when they “shouldn’t.”

Why quantum tunneling?

The discovery of quantum tunneling in the early 20th century allowed scientists to explain natural phenomena such as radioactive decay that seemed impossible according to classical physics.

In the 21st century, scientists face a similar problem. Existing theories fail to explain human perception, behavior, and decision-making.

Research has shown that the tools of quantum mechanics can help explain human behavior and decision-making.

Some have suggested that quantum effects play an important role in our brains. Even if this is not the case, we can still find the laws of quantum mechanics useful for modeling human thought. For example, quantum computing algorithms are more efficient than classical algorithms for many tasks.

With this in mind, I wanted to find out what would happen if I injected quantum effects into the functioning of a neural network.

So how does the quantum tunnel network work?

When we see an optical illusion with two possible interpretations (like the ambiguous cube or the vase and faces), researchers believe we temporarily hold both interpretations at the same time, until our brain decides which image should be seen.

This situation resembles the quantum thought experiment of Schrödinger’s cat. This famous scenario describes a cat in a box whose life depends on the decay of a quantum particle. According to quantum mechanics, the particle can be in two different states at the same time until we observe it. The cat can therefore be simultaneously alive and dead.

I trained my quantum tunneling neural network to recognize the Necker cube and Rubin vase illusions. Given the illusion as input, it produced an output corresponding to either interpretation.

Over time, the chosen interpretation oscillated between the two. Traditional neural networks also produce this behavior, but in addition, my network produced ambiguous results oscillating between the two certain outputs, just as our own brains can hold the two interpretations together before settling on one.

And now ?

In the age of deepfakes and fake news, understanding how our brains process illusions and construct models of reality has never been more important.

In other research, I explore how quantum effects can also help us understand social behavior and opinion radicalization in social networks.

In the long term, quantum AI could eventually contribute to the development of conscious robots. But for now, my research continues.

This article is republished from The Conversation under a Creative Commons license. Read the original article.