Optical illusions, quantum mechanics and neural networks appear to be quite different topics at first glance. latest research I used a phenomenon called “quantum tunneling” to design a neural network that may “see” optical illusions in an analogous technique to humans.
My neural network has simulated the human perception of the famous Necker Cube And Rubin's Vase Illusions – and truly higher than some much larger traditional neural networks utilized in computer vision.
This work could also make clear the query of whether artificial intelligence (AI) systems can ever truly achieve anything like human perception.
Ivan Maksymov
Why optical illusions?
Optical illusions deceive our brain into seeing things which will or will not be real. We don’t fully understand how optical illusions workbut studying them can teach us something about how our brain works and why it sometimes fails, for instance in cases like dementia and further long space flights.
Researchers using AI to mimic and study human vision have found that optical illusions are an issue. While computer vision systems can recognize complex objects, comparable to Art paintingsthey often can’t understand optical illusions. (The latest models at the least seem to acknowledge some kinds of illusionsbut these results require further investigation.)
My research addresses this problem using quantum physics.
How does my neural network work?
When the human brain processes information, it decides which data is useful and which just isn’t. A neural network mimics the function of the brain by utilizing many layers of artificial neurons that allow it to store data and classify it as useful or not.
Neurons are activated by signals from their neighbors. Imagine that every neuron has to climb over a brick wall to change into activated, and signals from neighbors push it higher and better until it finally climbs over the wall and reaches the activation point on the opposite side.
In quantum mechanics, tiny objects like electrons can sometimes penetrate seemingly impenetrable barriers. This happens through an effect called “quantum tunneling.” In my neural network, quantum tunneling sometimes allows neurons to leap all the way through the wall to the activation point and switch on, even after they “shouldn't.”
Why quantum tunneling?
The discovery of the quantum tunneling effect in the primary a long time of the twentieth century enabled scientists to clarify natural phenomena comparable to radioactive decay that seemed unattainable based on classical physics.
In the twenty first century, scientists face an analogous problem. Existing theories are insufficient to clarify human perception, behavior, and decision-making.
Research has shown that quantum mechanics tools can assist explain human behavior And Decision making.
While some suspect that quantum effects play a vital role in our brainEven if this just isn’t the case, we are able to still consider the laws of quantum mechanics useful for modeling human thought. For example, quantum algorithms are more efficient than classical algorithms for a lot of tasks.
With this in mind, I wanted to search out out what happens once I introduce quantum effects into the functioning of a neural network.
So how does the quantum tunneling network work?
When we see an optical illusion with two possible interpretations (comparable to the ambiguous cube or the vase and the faces), researchers imagine that we temporarily each interpretations concurrentlyuntil our brain decides which image needs to be seen.
This situation is comparable to the quantum mechanical thought experiment of Schrödinger's cat. This famous scenario describes a cat in a box whose life is dependent upon the decay of a quantum particle. According to quantum mechanics, the particle will be in two different states at the identical time until it’s observed – and so the cat will also be alive and dead at the identical time.
I trained my quantum tunneling neural network to acknowledge the Necker cube and Rubin's vase illusions. When presented with the illusion as input, it produced one or the opposite of the 2 interpretations as output.
Over time, the chosen interpretation fluctuated backwards and forwards. Traditional neural networks also produce this behavior, but as well as, my network produced some ambiguous results that fluctuated between the 2 particular outputs – just like how our own brains can hold each interpretations together before deciding on one.
What now?
In an era of Deepfakes – Translation And False reportsIt has never been more necessary to grasp how our brain processes illusions and creates models of reality.
In other research, I investigate how quantum effects also can help us understand Social behavior And Radicalization of opinions in social networks.
In the long run Quantum-based AI can ultimately contribute to conscious robots. But for now, my research work continues.
