You are looking at the image of a black circle on a grid of circular dots. It looks like a hole etched in a piece of white mesh fabric, although it’s actually a flat, fixed image on a screen or piece of paper. But your brain doesn’t understand it that way. Like a low-level hallucinatory experience, your mind stumbles; perceive the static image as the mouth of a dark tunnel heading towards you.
Responding to the likelihood of the effect, the body begins to react unconsciously: the pupils of the eye dilate to let in more light, just as they would adjust if you were about to be plunged into the darkness to ensure the best possible vision.
The effect in question was created by Akiyoshi Kitaoka, a psychologist at Ritsumeikan University in Kobe, Japan. It’s one of dozens of optical illusions he’s created over a long career. (“I love them all,” he said, answering Digital Trend’s question about whether he had a favorite.)
This new illusion was the subject of recently published research. in the journal Frontiers in Human Neuroscience. Although the article focuses firmly on human physiological responses to the new effect (which 86% of us will experience), the general topic may also have a lot of relevance when it comes to the future of artificial intelligence – like one of the researchers was quick to explain it to Digital Trends.
An evolutionary advantage
Table of Contents
Something is wrong with your brain. At least that’s an easy conclusion to draw from how the human brain perceives optical illusions. What other explanation is there for a static two-dimensional image that the brain perceives as something totally different? For a long time, mainstream psychology thought exactly that.
“Initially, people thought, ‘Okay, our brain isn’t perfect… It doesn’t always work properly.’ It’s a failure, isn’t it?” said Bruno only, a professor in the Department of Psychology at the University of Oslo and first author of the aforementioned study. “The illusions in this case were interesting because they revealed some sort of imperfection in the machinery.”
The brain has no way of knowing what is [really] the low.”
Psychologists don’t see them that way anymore. On the contrary, research like this highlights the fact that the visual system is not just a simple camera. The optical illusion “Illusory Expanding Hole” clearly indicates that the eye adapts to perceived, even imagined, light and dark, rather than physical energy.
More importantly, it shows that we are not simply recording the world with our visual systems, but rather performing an ongoing set of scientific experiments in order to gain a slight evolutionary advantage. The goal is to analyze the data presented to us and try to preemptively address issues before they become, well, problems.
“The brain has no way of knowing what is [really] there,” Laeng said. “What it does is build a sort of virtual reality of what might be out there. There is a bit of guesswork. In this regard, you can think of the brain as a sort of probabilistic machine. You can call it a Bayesian machine if you want. It uses a prior hypothesis and tries to test it all the time to see if it works.
Laeng gives the example of our eyes making adjustments based on nothing more than the impression of sunlight: even when seen through cloud cover or a canopy of leaves. In case.
“What matters in evolution is not that it’s true [at that moment], but it’s likely,” he continued. “By constricting the pupil, your body is already adapting to a situation that is very likely to occur in a short period of time. What happens [if the sun suddenly comes out] is that you are dazzled. Dazzled means temporarily unable. This has enormous consequences whether you are prey or predator. You lose a fraction of a second in a particular situation and you risk not surviving.
It’s not just light and dark where our visual systems have to make guesses, either. Think of a game of tennis, where the ball is moving at high speed. If we based our behavior entirely on what the visual system is receiving at any given time, we would fall behind reality and fail to pass the ball. “We are able to perceive the present although we are really stuck in the past,” Laeng said. “The only way to do that is to predict the future. It sounds a bit like a pun, but that’s it in a nutshell.
Machine vision improves
So what does this have to do with computer vision? Potentially everything. For a robot, for example, to operate effectively in the real world, it must be able to make these kinds of adjustments on the fly. Computers have an advantage when it comes to their ability to perform extremely fast calculations. What they don’t have are millions of years of evolution on their side.
In recent years, artificial vision has nevertheless made enormous progress. They can identify faces or gaits in real-time video streams, potentially even in vast crowds of people. Similar image classification and technology tools can also recognize the presence of other objects, while breakthroughs in object segmentation provide insight into the content of different scenes. Significant progress has also been made in extrapolating 3D images from 2D scenes, allowing machines to “read” three-dimensional information, such as depth, from scenes. This brings modern computer vision closer to human image perception.
However, there is still a gap between the best machine vision algorithms and the kinds of vision-based abilities that the overwhelming majority of humans are able to achieve from a young age. While we can’t articulate exactly how we perform these vision-based tasks (to quote Hungarian-British polymath Michael Polanyi, “we can know more than we can tell”), we are nonetheless able to perform an impressive array of tasks that allow us to exploit our eyesight in a variety of clever ways.
A Turing test for machine vision
If researchers and engineers hope to create computer vision systems that perform at least on par with the visual processing capabilities of the wetware brain, building algorithms that can understand optical illusions isn’t a bad thing. starting point. At the very least, this could prove to be a good way to measure how well machine vision systems work with our own brains. This may not be the answer to the mythical General Artificial Intelligencebut it could be the key to unlocking General Vision.
“If someone ever developed an artificial visual system that makes the same illusory errors of perception that we do, you would know at this point that it is [achieving] a good simulation of how our brain works,” Laeng said. “It would be a kind of Turing test. If you have an artificial network deceived by illusion like we are, then we [would be] very close to understanding the underlying computation of the brain itself.
Song Yi-Zhe, reader of Computer Vision and Machine Learning at the Center for Vision Speech and Signal Processing at the University of Surrey in the UK, agrees with the hypothesis. “Asking vision algorithms to understand optical illusions as a general topic is of great value to the community,” he told Digital Trends. “This goes beyond the current community goal of asking machines to [recognize]by pushing the envelope further [and] ask the machines to reason. This push [would represent] a significant step forward towards the “big picture”, where subjective interpretations of visual concepts must be taken into account.
Use your illusion
To date, there has been some limited research towards this goal – although it remains at a relatively early stage. Nasim Nematzadeh, PhD researcher. in artificial intelligence and robotics – low-level vision models, is a person who has published works on this topic.
“We believe that further exploration of the role of simple Gaussian-like patterns in low-level retinal and early-stage Gaussian nucleus processing [deep neural networks], and its prediction of perceptual illusion loss, will lead to more accurate computer vision techniques and models,” Nematzadeh told Digital Trends. “[This could] contribute to higher-level models of depth and motion processing and generalized to computational understanding of natural images.
Max Williams, an AI researcher who helped compile a dataset of thousands of optical illusion pictures for computer vision systems, presents the relationship between general vision and optical illusions most succinctly: “Illusions exist because our eyes and brain perform a messy, ad hoc process to extract a scene visual of an otherwise incomprehensible field of light, created by a world from which we are almost completely isolated,” they told Digital Trends. “I don’t think it’s possible to make a visual system expressive enough to be considered a a “perception” that is also free from illusions.”
Achieve the overall vision
To be clear, achieving general human-level (or better) vision for the AI isn’t just going to train them to recognize standard optical illusions. No amount of hyper-specific ability to, say, decode Magic Eye illusions with 99.9% accuracy in 0.001 seconds is going to substitute for millions of years of human evolution.
(Interestingly, computer vision already has its own version of optical illusions in the form of conflicting patterns, which can cause them to go wrong – as in an alarming illustration – a 3D printed toy turtle for a rifle. However, these do not give the same evolutionary benefits as the optical illusions that work on humans.)
Still, getting machines to understand and respond to human optical illusions the way we do could be very useful research.
And one thing is certain: when General Vision AI is reached, he will fall into the same kind of optical illusions as we do. At least, in the case of the illusory expanding hole, 86% of us.
Editors’ Recommendations