Images that humans perceive as completely unrelated can be classified as the same by computational models.
Human sensory systems are very good at recognizing objects that we see or words that we hear, even if the object is upside down or the word is spoken by a voice we’ve never heard.
Computational models known as deep neural networks can be trained to do the same thing, correctly identifying an image of a dog regardless of what color its fur is, or a word regardless of the pitch of the speaker’s voice. However, a new study from MIT neuroscientists has found that these models often also respond the same way to images or words that have no resemblance to the target.
When these neural networks were used to generate an image or a word that they responded to in the same way as a specific natural input, such as a picture of a bear, most of them generated images or sounds that were unrecognizable to human observers. This suggests that these models build up their own idiosyncratic “invariances” — meaning that they respond the same way to stimuli with very different features.
The findings offer a new way for researchers to evaluate how well these models mimic the organization of human sensory perception, says Josh McDermott, an associate professor of brain and cognitive sciences at MIT and a member of MIT’s McGovern Institute for Brain Research and Center for Brains, Minds, and Machines.
“This paper shows that you can use these models to derive unnatural signals that end up being very diagnostic of the representations in the model,” says McDermott, who is the senior author of the study. “This test should become part of a battery of tests that we as a field are using to evaluate models.”
Jenelle Feather PhD ’22, who is now a research fellow at the Flatiron Institute Center for Computational Neuroscience, is the lead author of the open-access paper, which appears today in Nature Neuroscience. Guillaume Leclerc, an MIT graduate student, and Aleksander Mądry, the Cadence Design Systems Professor of Computing at MIT, are also authors of the paper.
Different perceptions
In recent years, researchers have trained deep neural networks that can analyze millions of inputs (sounds or images) and learn common features that allow them to classify a target word or object roughly as accurately as humans do. These models are currently regarded as the leading models of biological sensory systems.
It is believed that when the human sensory system performs this kind of classification, it learns to disregard features that aren’t relevant to an object’s core identity, such as how much light is shining on it or what angle it’s being viewed from. This is known as invariance, meaning that objects are perceived to be the same even if they show differences in those less important features.
“Classically, the way that we have thought about sensory systems is that they build up invariances to all those sources of variation that different examples of the same thing can have,” Feather says. “An organism has to recognize that they’re the same thing even though they show up as very different sensory signals.”
The researchers wondered if deep neural networks that are trained to perform classification tasks might develop similar invariances. To try to answer that question, they used these models to generate stimuli that produce the same kind of response within the model as an example stimulus given to the model by the researchers.
They term these stimuli “model metamers,” reviving an idea from classical perception research whereby stimuli that are indistinguishable to a system can be used to diagnose its invariances. The concept of metamers was originally developed in the study of human perception to describe colors that look identical even though they are made up of different wavelengths of light.
To their surprise, the researchers found that most of the images and sounds produced in this way looked and sounded nothing like the examples that the models were originally given. Most of the images were a jumble of random-looking pixels, and the sounds resembled unintelligible noise. When researchers showed the images to human observers, in most cases the humans did not classify the images synthesized by the models in the same category as the original target example.
“They’re really not recognizable at all by humans. They don’t look or sound natural and they don’t have interpretable features that a person could use to classify an object or word,” Feather says.
The findings suggest that the models have somehow developed their own invariances that are different from those found in human perceptual systems. This causes the models to perceive pairs of stimuli as being the same despite their being wildly different to a human.
Idiosyncratic invariances
The researchers found the same effect across many different vision and auditory models. However, each of these models appeared to develop their own unique invariances. When metamers from one model were shown to another model, the metamers were just as unrecognizable to the second model as they were to human observers.
“The key inference from that is that these models seem to have what we call idiosyncratic invariances,” McDermott says. “They have learned to be invariant to these particular dimensions in the stimulus space, and it’s model-specific, so other models don’t have those same invariances.”
The researchers also found that they could induce a model’s metamers to be more recognizable to humans by using an approach called adversarial training. This approach was originally developed to combat another limitation of object recognition models, which is that introducing tiny, almost imperceptible changes to an image can cause the model to misrecognize it.
The researchers found that adversarial training, which involves including some of these slightly altered images in the training data, yielded models whose metamers were more recognizable to humans, though they were still not as recognizable as the original stimuli. This improvement appears to be independent of the training’s effect on the models’ ability to resist adversarial attacks, the researchers say.
“This particular form of training has a big effect, but we don’t really know why it has that effect,” Feather says. “That’s an area for future research.”
Analyzing the metamers produced by computational models could be a useful tool to help evaluate how closely a computational model mimics the underlying organization of human sensory perception systems, the researchers say.
“This is a behavioral test that you can run on a given model to see whether the invariances are shared between the model and human observers,” Feather says. “It could also be used to evaluate how idiosyncratic the invariances are within a given model, which could help uncover potential ways to improve our models in the future.”
Reference: “Model metamers reveal divergent invariances between biological and artificial neural networks” by Jenelle Feather, Guillaume Leclerc, Aleksander Mądry and Josh H. McDermott, 16 October 2023, Nature Neuroscience. DOI: 10.1038/s41593-023-01442-0
The research was funded by the National Science Foundation, the National Institutes of Health, a Department of Energy Computational Science Graduate Fellowship, and a Friends of the McGovern Institute Fellowship.
News
Are Machines Truly Thinking? Modern AI Systems Have Finally Achieved Turing’s Vision
Modern AI systems have fulfilled Turing’s vision of machines that learn and converse like humans, but challenges remain. A new paper highlights concerns about energy consumption and societal inequality while calling for more robust [...]
The Surprising Link Between Smell, Sound, and Emotions
New research reveals how smell and hearing interact in the brain to drive social behavior, using mouse maternal instincts as a model. Imagine you’re at a dinner party, but you can’t smell the food [...]
Brain cells age at different rates
As our body ages, not only joints, bones and muscles wear out, but also our nervous system. Nerve cells die, are no longer fully replaced, and the brain shrinks. "Aging is the most important risk factor [...]
Long COVID Breakthrough: Spike Proteins Persist in Brain for Years
Researchers have discovered that the SARS-CoV-2 spike protein persists in the brain and skull bone marrow for years after infection, potentially leading to chronic inflammation and neurodegenerative diseases. Researchers from Helmholtz Munich and Ludwig-Maximilians-Universität (LMU) have [...]
Water-Resistant Paper Could Revolutionize Packaging and Replace Plastic
A groundbreaking study showcases the creation of sustainable hydrophobic paper, enhanced by cellulose nanofibres and peptides, presenting a biodegradable alternative to petroleum-based materials, with potential uses in packaging and biomedical devices. Researchers aimed to [...]
NIH Scientists Discover Game-Changing Antibodies Against Malaria
Novel antibodies have the potential to pave the way for the next generation of malaria interventions. Researchers at the National Institutes of Health (NIH) have identified a novel class of antibodies that target a previously unexplored region [...]
Surprising Discovery: What If Some Cancer Genes Are Actually Protecting You?
A surprising discovery reveals that a gene previously thought to accelerate esophageal cancer actually helps protect against it initially. This pivotal study could lead to better prediction and prevention strategies tailored to individual genetic [...]
The Cancer Test That Exposes What Conventional Scans Miss
Researchers at UCLA have unveiled startling findings using PSMA-PET imaging that reveal nearly half of patients diagnosed with high-risk prostate cancer might actually have metastases missed by traditional imaging methods. This revelation could profoundly affect future [...]
Pupil size in sleep reveals how memories are processed
Cornell University researchers have found that the pupil is key to understanding how, and when, the brain forms strong, long-lasting memories. By studying mice equipped with brain electrodes and tiny eye-tracking cameras, the researchers [...]
Stanford’s Vaccine Breakthrough Boosts Flu Protection Like Never Before
Stanford Medicine researchers have developed a new method for influenza vaccination that encourages a robust immune response to all four common flu subtypes, potentially increasing the vaccine’s efficacy. In laboratory tests using human tonsil [...]
Water’s Worst Nightmare: The Rise of Superhydrophobic Materials
New materials with near-perfect water repellency offer potential for self-cleaning surfaces in cars and buildings. Scientists from Karlsruhe Institute of Technology (KIT) and the Indian Institute of Technology Guwahati (IITG) have developed a surface [...]
Japanese dentists test drug to help people with missing teeth regrow new ones
Japanese dentists are testing a groundbreaking drug that could enable people with missing teeth to grow new ones, reducing the need for dentures and implants, AFP recently reported. Katsu Takahashi, head of oral surgery at [...]
An AI system has reached human level on a test for ‘general intelligence’
A new artificial intelligence (AI) model has just achieved human-level results on a test designed to measure "general intelligence." On December 20, OpenAI's o3 system scored 85% on the ARC-AGI benchmark, well above the previous AI best [...]
According to Researchers, Your Breathing Patterns Could Hold the Key to Better Memory
Breathing synchronizes brain waves that support memory consolidation. A new study from Northwestern Medicine reports that, much like a conductor harmonizes various instruments in an orchestra to create a symphony, breathing synchronizes hippocampal brain waves to [...]
The Hidden Culprit Behind Alzheimer’s Revealed: Microglia Under the Microscope
Researchers at the CUNY Graduate Center have made a groundbreaking discovery in Alzheimer’s disease research, identifying a critical link between cellular stress in the brain and disease progression. Their study focuses on microglia, the brain’s immune [...]
“Mirror Bacteria” Warning: A New Kind of Life Could Pose a Global Threat
Mirror life, a concept involving synthetic organisms with reversed molecular structures, carries significant risks despite its potential for medical advancements. Experts warn that mirror bacteria could escape natural biological controls, potentially evolving to exploit [...]