No CrossRef data available.
Article contents
There is a fundamental, unbridgeable gap between DNNs and the visual cortex
Published online by Cambridge University Press: 06 December 2023
Abstract
Deep neural networks (DNNs) are not just inadequate models of the visual system but are so different in their structure and functionality that they are not even on the same playing field. DNN units have almost nothing in common with neurons, and, unlike visual neurons, they are often fully connected. At best, DNNs can label inputs, while our object perception is both holistic and detail preserving. A feat that no computational system can achieve.
- Type
- Open Peer Commentary
- Information
- Copyright
- Copyright © The Author(s), 2023. Published by Cambridge University Press
References
Crouzet, S. M., Kirchner, H., & Thorpe, S. J. (2010). Fast saccades toward faces: Face detection in just 100 ms. Journal of Vision, 10, 10–17.CrossRefGoogle Scholar
Freiwald, W. A., & Tsao, D. Y. (2010). Functional compartmentalization and viewpoint generalization within the macaque face processing system. Science (New York, N.Y.), 330, 845–851.CrossRefGoogle ScholarPubMed
Greene, E., & Visani, A. (2015). Recognition of letters displayed as briefly flashed dot patterns. Attention Perception & Psychophysics, 77, 1955–1969.CrossRefGoogle ScholarPubMed
Gur, M. (2018). Very small faces are easily discriminated under long and short exposure times. Journal of Neurophysiology, 119, 1599–1607. doi:10.1152/jn.00622.2017CrossRefGoogle ScholarPubMed
Gur, M. (2021). Psychophysical evidence and perceptual observations show that object recognition is not hierarchical but is a parallel, simultaneous, egalitarian, non-computational system. BioRxiv, 1–25. doi:10.1101/2021.06.10.447325Google Scholar
Keysers, C., Xiao, D.-K., Foldiak, P., & Perrett, D. I. (2001). The speed of light. Journal of Cognitive Neuroscience, 13, 90–101.CrossRefGoogle Scholar
Rolls, E. T. (2012). Invariant visual object and face recognition: Neural and computational bases, and a model, VisNet. Frontiers in Computational Neuroscience, 6, 1–70.CrossRefGoogle Scholar
Schmolesky, M. T., Youngchang, W., Hanes, D. P., Thompson, K. G., Leutgeb, S., Schall, J. D., & Leventhal, A. G. (1998). Signal timing in the macaque visual system. Journal of Neurophysiology, 79, 3272–3278.CrossRefGoogle ScholarPubMed
Stettler, D. D., Aniruddha, D., Bennett, J., & Gilbert, C. D. (2002). Lateral connectivity and contextual interactions in macaque primary visual cortex. Neuron, 36, 739–750.CrossRefGoogle ScholarPubMed
Ullman, S. (2019). Using neuroscience to develop artificial intelligence. Science (New York, N.Y.), 363, 692–693.CrossRefGoogle ScholarPubMed
Target article
Deep problems with neural network models of human vision
Related commentaries (29)
Explananda and explanantia in deep neural network models of neurological network functions
A deep new look at color
Beyond the limitations of any imaginable mechanism: Large language models and psycholinguistics
Comprehensive assessment methods are key to progress in deep learning
Deep neural networks are not a single hypothesis but a language for expressing computational hypotheses
Even deeper problems with neural network models of language
Fixing the problems of deep neural networks will require better training data and learning algorithms
For deep networks, the whole equals the sum of the parts
For human-like models, train on human-like tasks
Going after the bigger picture: Using high-capacity models to understand mind and brain
Implications of capacity-limited, generative models for human vision
Let's move forward: Image-computable models and a common model evaluation scheme are prerequisites for a scientific understanding of human vision
Modelling human vision needs to account for subjective experience
Models of vision need some action
My pet pig won't fly and I want a refund
Neither hype nor gloom do DNNs justice
Neural networks need real-world behavior
Neural networks, AI, and the goals of modeling
Perceptual learning in humans: An active, top-down-guided process
Psychophysics may be the game-changer for deep neural networks (DNNs) to imitate the human vision
Statistical prediction alone cannot identify good models of behavior
The model-resistant richness of human visual experience
The scientific value of explanation and prediction
There is a fundamental, unbridgeable gap between DNNs and the visual cortex
Thinking beyond the ventral stream: Comment on Bowers et al.
Using DNNs to understand the primate vision: A shortcut or a distraction?
Where do the hypotheses come from? Data-driven learning in science and the brain
Why psychologists should embrace rather than abandon DNNs
You can't play 20 questions with nature and win redux
Author response
Clarifying status of DNNs as models of human vision