pith. sign in

arxiv: 2606.12340 · v1 · pith:2XWPGHLOnew · submitted 2026-06-10 · 💻 cs.CV

Echoes of the Prior: A Computational Phenomenology of Forgetting

Gege Gao , Bernhard Sch\"olkopf , Andreas Geiger This is my paper

Pith reviewed 2026-06-27 09:49 UTC · model grok-4.3

classification 💻 cs.CV
keywords forgettingsynaptic decay3D reconstructionphenomenologyinteractive installationpredictive priorsneuromorphic aestheticsneural network degradation
0
0 comments X

The pith

Controlled decay in a feed-forward 3D reconstruction network generates visual chaos that stands in for the experience of human forgetting.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper builds an interactive installation around a 3D reconstruction model whose connections are deliberately weakened over time. As the network loses its ability to maintain consistent scene structure, the output shifts from recognizable environments into fragmented and unrecognizable forms. This degradation is offered as a direct computational stand-in for the way biological memory loss removes the predictive scaffolding that keeps everyday perception coherent. A sympathetic reader would see value in having a runnable silicon example of how forgetting does not simply erase data but actively dissolves the world into disorientation. The work treats the neural network itself as the central object whose structural erosion produces the intended subjective effect.

Core claim

By applying controlled synaptic decay inside a feed-forward 3D reconstruction model, the installation produces a sequence of outputs that regress from coherent scenes into unrecognizable chaos, thereby creating an artistic analogy for the erosion of the brain's predictive priors and positioning the neural network as a cognitive proxy for the disorienting experience of losing one's grip on the world.

What carries the argument

Feed-forward 3D reconstruction model with induced controlled synaptic decay, which progressively removes the network's capacity to enforce consistent scene geometry.

If this is right

  • The same decay process can be applied to other reconstruction or generative models to produce analogous visualizations of cognitive erosion.
  • The installation supplies a concrete, runnable example that the community can extend for further neuromorphic aesthetic experiments.
  • If the analogy holds, it supplies a method for visualizing how loss of predictive structure turns ordered perception into chaos without requiring biological subjects.
  • The framework treats neural networks as proxies rather than tools, opening repeated use of degradation as an artistic and exploratory device.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Similar decay schedules could be tested on models trained for other modalities such as audio or language to check whether the same progression toward incoherence appears.
  • The installation might serve as a prompt for psychological experiments that compare reported subjective states during real forgetting with reported states during exposure to the decayed outputs.
  • If the visual progression reliably produces the intended effect, the technique could be adapted to simulate early-stage cognitive decline for educational or therapeutic visualization purposes.

Load-bearing premise

The visual output of a degrading neural network can meaningfully evoke or stand in for the subjective, disorienting phenomenology of human forgetting.

What would settle it

Viewers who interact with the installation and report no increase in disorientation or sense of lost coherence as the model decays would falsify the central analogy.

Figures

Figures reproduced from arXiv: 2606.12340 by Andreas Geiger, Bernhard Sch\"olkopf, Gege Gao.

Figure 1
Figure 1. Figure 1: Echoes of the Prior. A computational simulation of the fragility of memory under entropic loss. Memory is not merely the storage of data; it is the scaffolding of reality. When biological memory fades, the world does not simply turn black; it regresses into an unrecognizable chaos. Echoes of the Prior is an interactive installation that attempts to visualize this subjective phenomenology of forgetting. By … view at source ↗
Figure 2
Figure 2. Figure 2: Installation Design of Echoes of the Prior. Snapshot of the real-time feedback (𝜆 = 0.21). The viewport displays the immediate structural dissolution of the 3D scene as the user adjusts the decay factor. 1 Introduction: The Invisible Process “What I cannot create, I do not understand” is a famous quote by physicist Richard Feynman. While neuroscience can explain the mechanism of Alzheimer’s – the plaques, … view at source ↗
Figure 3
Figure 3. Figure 3: t-SNE visualization of the DINOv2 latent space. [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Simulating Sensory Decay Gsense. (Left) Healthy Baseline (𝜆 = 0): Response of a standard Gabor filter (V1 primitive) to a sharp stimulus. (Right) Pathology: As 𝜆 increases, the filter degrades via three mechanisms: (i) Spectral Amnesia (loss of high-frequency acuity); (ii) Neural Fabrication (injection of coherent noise/hallucinations); and (iii) Structural Atrophy (synaptic sparsity leading to signal loss… view at source ↗
Figure 5
Figure 5. Figure 5: Topology of Memory Decay. (a) Manifold Twist: Visualization of the orthogonal rotation R(𝜆). As 𝜆 increases (blue → purple), the feature vector rotates away from its semantic alignment while preserving its norm. (b) Computational Uncanny Valley: A computationally defined model of visual Uncanniness (purple), defined as the product of structural integrity (norm) and semantic distortion (1 − cosine similarit… view at source ↗
Figure 6
Figure 6. Figure 6: Visualizing Selective Forgetting. Targeted entropy injection using a text prompt ("woman in red"). (Left) Original reconstruction. (Right) Decayed result (𝜆 = 0.6). The targeted entity undergoes topological dissolution, effectively ghosting out of reality while the surrounding context remains intact. 3.3.2 Latent Space Surgery. To extract the prior about the input images, the vision foundation model [Oquab… view at source ↗
Figure 7
Figure 7. Figure 7: The System Workflow of Echoes of the Prior. (see Sec. 3.2.1); The Memory Decay Path (Bottom), which optionally integrates a Selective Forgetting module. This module utilizes Grounded-SAM [Ren et al. 2024] to generate semantic masks based on user prompts. Entropy is surgically injected only into the tokens representing specific objects, disrupting their structural integrity while preserving the context. Fin… view at source ↗
Figure 8
Figure 8. Figure 8: The Interactive Interface. The web-based GUI enables real-time interaction with the decay process. Proc. ACM Comput. Graph. Interact. Tech., Vol. 9, No. 3, Article 42. Publication date: July 2026 [PITH_FULL_IMAGE:figures/full_fig_p010_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: “I Glitch, Ergo Sum” (I glitch, therefore I am). A visual manifesto generated by our real-time streaming system DecArt. 4.3 The Experience Journey We describe the intended audience encounter to articulate the experiential arc of the installation. Encounter and Orientation. The viewer enters a darkened space and approaches a large display showing a photorealistic 3D scene, with a tablet serving as the contr… view at source ↗
Figure 10
Figure 10. Figure 10: Memory as a Potemkin Village: Hallucination vs. Reconstruction. While 3D generative models (Right) fabricate a complete, watertight object by inventing unseen details, our system (Left) preserves the epistemological honesty of the input. It projects a thin, fragile film, like a facade of memory, that disintegrates when the viewer attempts to look behind the veil. This hollowness is not a technical failure… view at source ↗
read the original abstract

Memory is not merely the storage of data; it is the scaffolding of reality. When biological memory fades, the world does not simply turn black; it regresses into an unrecognizable chaos. Echoes of the Prior is an interactive installation that attempts to visualize this subjective phenomenology of forgetting. By inducing controlled synaptic decay within a Feed-Forward 3D Reconstruction model, we create an artistic analogy for the erosion of the brain's predictive priors. We position the Neural Network not as a tool for engineering, but as a cognitive proxy - a silicon brain whose structural degeneration evokes the disorienting, poetic, and terrifying experience of losing one's grip on the world. Ultimately, we offer this framework as a catalyst, inviting the wider community to explore the uncharted potential of neuromorphic aesthetics in visualizing the fragility of intelligence. Interactive demo see https://decart-4d.github.io/.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

0 major / 2 minor

Summary. The manuscript presents 'Echoes of the Prior,' an interactive installation that induces controlled synaptic decay in a Feed-Forward 3D Reconstruction model to create an artistic analogy for the erosion of the brain's predictive priors during forgetting. It frames the neural network as a cognitive proxy whose structural degeneration evokes the subjective, disorienting phenomenology of memory loss, and invites exploration of neuromorphic aesthetics without asserting biological or empirical equivalence.

Significance. If the intended analogy holds as an artistic device, the work contributes to neuromorphic aesthetics by offering a visual and interactive framework for representing cognitive fragility. Its explicit positioning as an analogy and catalyst for community exploration, rather than a mechanistic model, is a strength that aligns with the stated goals and distinguishes it from technical claims in computer vision.

minor comments (2)
  1. [Abstract] Abstract: the phrase 'controlled synaptic decay' is introduced without any accompanying description of the decay schedule, parameterization, or implementation details that would allow readers to understand or replicate the visual effects.
  2. The manuscript provides a demo link but contains no description of the interaction mechanics, input modalities, or how users are expected to experience the degradation process.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for their positive assessment of the manuscript and their recommendation to accept. We appreciate the recognition that the work is framed explicitly as an artistic analogy rather than a mechanistic or empirical model, which aligns with our intent to contribute to neuromorphic aesthetics.

Circularity Check

0 steps flagged

No derivation chain or quantitative claims; artistic analogy only

full rationale

The paper describes an interactive artistic installation that uses controlled degradation of a feed-forward 3D reconstruction network as a visual proxy for the phenomenology of forgetting. It makes no equations, predictions, fitted parameters, or load-bearing derivations of any kind. The central claim is explicitly positioned as an analogy and aesthetic exploration rather than a mechanistic or empirical result. No self-citations, ansatzes, or reductions to inputs exist that could trigger any of the enumerated circularity patterns. The work is self-contained within its stated artistic goals.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review; no model equations, parameters, or new entities are described.

pith-pipeline@v0.9.1-grok · 5680 in / 953 out tokens · 15648 ms · 2026-06-27T09:49:55.538126+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

40 extracted references · 3 linked inside Pith

  1. [1]

    InACM SIGGRAPH 2019 Art Gallery

    Learning to see: you are what you see. InACM SIGGRAPH 2019 Art Gallery. 1–6. Refik Anadol

    2019

  2. [2]

    Anne Davis Basting

    Working memory.Science255, 5044 (1992), 556–559. Anne Davis Basting. 2009.Forget Memory: Creating Better Lives for People with Dementia. Johns Hopkins University Press. Jeffrey S. Bowers, Gaurav Malhotra, Marin Dujmović, Milton Llera Montero, Christian Tsvetkov, Valerio Biscione, Guillermo Puebla, Federico Adolfi, John E. Hummel, Rachel F. Heaton, and et al

    1992

  3. [3]

    David J Chalmers

    Deep problems with neural network models of human vision.Behavioral and Brain Sciences46 (2023), e385. David J Chalmers

    2023

  4. [4]

    Radoslaw Martin Cichy and Daniel Kaiser

    Facing up to the problem of consciousness.Journal of Consciousness Studies2, 3 (1995), 200–219. Radoslaw Martin Cichy and Daniel Kaiser

    1995

  5. [5]

    Philip R Corlett, Guillermo Horga, Paul C Fletcher, Ben Alderson-Day, Katharina Schmack, and Albert R Powers

    Deep neural networks as scientific models.Trends in Cognitive Sciences23, 4 (2019), 305–317. Philip R Corlett, Guillermo Horga, Paul C Fletcher, Ben Alderson-Day, Katharina Schmack, and Albert R Powers

    2019

  6. [6]

    Charles R

    Hallucinations and strong priors.Trends in cognitive sciences23, 2 (2019), 114–127. Charles R. Crutch, Sebastian J.and Harrison, Emilie V. Brotherhood, Paul M. Camic, Brian Day, and Anthony J. Woods

    2019

  7. [7]

    Sebastian J Crutch, Ron Isaacs, and Martin N Rossor

    Created Out of Mind: Shaping Perceptions of Dementia Through Art and Science.Mobile Brain-Body Imaging and the Neuroscience of Art, Innovation and Creativity(2019), 135–143. Sebastian J Crutch, Ron Isaacs, and Martin N Rossor

    2019

  8. [8]

    Russell L De Valois, Duane G Albrecht, and Lisa G Thorell

    Some workmen can blame their tools: artistic change in an individual with Alzheimer’s disease.The Lancet357, 9274 (07 2001), 2129–2133. Russell L De Valois, Duane G Albrecht, and Lisa G Thorell

    2001

  9. [9]

    C.H Espinel

    Spatial frequency selectivity of cells in macaque visual cortex.Vision research22, 5 (1982), 545–559. C.H Espinel

    1982

  10. [10]

    Martha J Farah

    de Kooning’s late colours and forms: dementia, creativity, and the healing power of art.The Lancet347, 9008 (1996), 1096–1098. Martha J Farah. 2004.Visual Agnosia. MIT Press. Sigmund Freud

    1996

  11. [11]

    Leon A Gatys, Alexander S Ecker, and Matthias Bethge

    The free-energy principle: a unified brain theory?Nature reviews neuroscience11, 2 (2010), 127–138. Leon A Gatys, Alexander S Ecker, and Matthias Bethge

    2010

  12. [12]

    Zehuan Huang, Yuan-Chen Guo, Xingqiao An, Yunhan Yang, Yangguang Li, Zi-Xin Zou, Ding Liang, Xihui Liu, Yan-Pei Cao, and Lu Sheng

    Neural networks and physical systems with emergent collective computational abilities.Proceedings of the National Academy of Sciences79, 8 (1982), 2554–2558. Zehuan Huang, Yuan-Chen Guo, Xingqiao An, Yunhan Yang, Yangguang Li, Zi-Xin Zou, Ding Liang, Xihui Liu, Yan-Pei Cao, and Lu Sheng

    1982

  13. [13]

    Eric Jonas and Konrad Paul Kording

    Receptive fields, binocular interaction and functional architecture in the cat’s visual cortex.The Journal of Physiology160, 1 (1962), 106–154. Eric Jonas and Konrad Paul Kording

    1962

  14. [14]

    Bernhard Kerbl, Georgios Kopanas, Thomas Leimkühler, and George Drettakis

    Could a neuroscientist understand a microprocessor?PLOS Computational Biology13, 1 (2017), 1–24. Bernhard Kerbl, Georgios Kopanas, Thomas Leimkühler, and George Drettakis

    2017

  15. [15]

    SIGGRAPH

    3d gaussian splatting for real-time radiance field rendering.ACM Transactions on Graphics (TOG)42, 4 (2023), 1–14. SIGGRAPH

    2023

  16. [16]

    1997.Dementia Reconsidered: The Person Comes First

    Tom Kitwood. 1997.Dementia Reconsidered: The Person Comes First. Open University Press. Pia Kontos

    1997

  17. [17]

    Shu-Chen Li, Ulman Lindenberger, and Sverker Sikström

    Embodied selfhood in Alzheimer’s disease: Rethinking person-centred care.Dementia: The International Journal of Social Research and Practice4 (2005), 553–570. Shu-Chen Li, Ulman Lindenberger, and Sverker Sikström

    2005

  18. [18]

    Trends in cognitive sciences5, 11 (2001), 479–486

    Aging cognition: from neuromodulation to representation. Trends in cognitive sciences5, 11 (2001), 479–486. Proc. ACM Comput. Graph. Interact. Tech., Vol. 9, No. 3, Article

    2001

  19. [19]

    arXiv:2511.10647 [cs.CV] Grace W

    Depth Anything 3: Recovering the Visual Space from Any Views. arXiv:2511.10647 [cs.CV] Grace W. Lindsay

    Pith/arXiv arXiv

  20. [20]

    Wolfgang Maass

    Convolutional neural networks as a model of the visual system: Past, present, and future.Journal of Cognitive Neuroscience33, 10 (2021), 2017–2031. Wolfgang Maass

    2021

  21. [21]

    Rosa Menkman

    Networks of spiking neurons: The third generation of neural network models.Neural networks10, 9 (1997), 1659–1671. Rosa Menkman. 2011.The glitch moment(um). Institute of Network Cultures. Maurice Merleau-Ponty. 1945.Phenomenology of Perception. Gallimard. Francesco Mezzadri

    1997

  22. [22]

    Ben Mildenhall, Pratul P Srinivasan, Matthew Tancik, Jonathan T Barron, Ravi Ramamoorthi, and Ren Ng

    How to generate random matrices from the classical compact groups.Notices of the AMS54, 5 (2007), 592–604. Ben Mildenhall, Pratul P Srinivasan, Matthew Tancik, Jonathan T Barron, Ravi Ramamoorthi, and Ren Ng

    2007

  23. [23]

    A Bayesian explanation of the ‘Uncanny Valley’ effect and related psychological phenomena.Scientific Reports2 (2012),

    2012

  24. [24]

    Fabian Offert and Peter Bell

    Inceptionism: Going deeper into neural networks.Google Research Blog(2015). Fabian Offert and Peter Bell

    2015

  25. [25]

    John O’Keefe and Jonathan Dostrovsky

    Perceptual bias and technical metapictures: Critical machine vision as a humanities challenge.AI & Society36 (2021), 1–12. John O’Keefe and Jonathan Dostrovsky

    2021

  26. [26]

    Nobel Prize winning work on Place Cells

    The hippocampus as a spatial map: preliminary evidence from unit activity in the freely-moving rat.Brain research34, 1 (1971), 171–175. Nobel Prize winning work on Place Cells. Maxime Oquab, Timothée Darcet, Théo Moutakanni, Huy V. Vo, Marc Szafraniec, Vasil Khalidov, Pierre Fernandez, Daniel HAZIZA, Francisco Massa, Alaaeldin El-Nouby, Mido Assran, Nicol...

    1971

  27. [27]

    Cynthia Owsley, Robert Sekuler, and Dennis Siemsen

    DINOv2: Learning Robust Visual Features without Supervision.Transactions on Machine Learning Research(2024). Cynthia Owsley, Robert Sekuler, and Dennis Siemsen

    2024

  28. [28]

    Albert R Powers, Christoph Mathys, and Philip R Corlett

    Contrast sensitivity throughout adulthood.Vision Research23, 7 (1983), 689–699. Albert R Powers, Christoph Mathys, and Philip R Corlett

    1983

  29. [29]

    René Ranftl, Alexey Bochkovskiy, and Vladlen Koltun

    Pavlovian conditioning-induced hallucinations result from overweighting of perceptual priors.Science357, 6351 (2017), 596–600. René Ranftl, Alexey Bochkovskiy, and Vladlen Koltun

    2017

  30. [30]

    Charles Bonnet Syndrome: Evidence for a Generative Model in the Cortex?PLOS Computational Biology9, 7 (2013), e1003134. Tianhe Ren, Shilong Liu, Ailing Zeng, Jing Lin, Kunchang Li, He Cao, Jiayu Chen, Xinyu Huang, Yukang Chen, Feng Yan, Zhaoyang Zeng, Hao Zhang, Feng Li, Jie Yang, Hongyang Li, Qing Jiang, and Lei Zhang

    2013

  31. [31]

    arXiv:2401.14159 [cs.CV] Jun’ichiro Seyama and Ruth S

    Grounded SAM: Assembling Open-World Models for Diverse Visual Tasks. arXiv:2401.14159 [cs.CV] Jun’ichiro Seyama and Ruth S. Nagayama

    Pith/arXiv arXiv

  32. [32]

    Michael N Shadlen and William T Newsome

    The Uncanny Valley: Effect of Realism on the Impression of Artificial Human Faces.Presence: Teleoperators and Virtual Environments16, 4 (2007), 337–351. Michael N Shadlen and William T Newsome

    2007

  33. [33]

    Oriane Siméoni, Huy V

    The Variable Discharge of Cortical Neurons: Implications for Connectivity, Computation, and Information Coding.Journal of Neuroscience18, 10 (1998), 3870–3896. Oriane Siméoni, Huy V. Vo, Maximilian Seitzer, Federico Baldassarre, Maxime Oquab, Cijo Jose, Vasil Khalidov, Marc Szafraniec, Seungeun Yi, Michaël Ramamonjisoa, Francisco Massa, Daniel Haziza, Luc...

    1998

  34. [34]

    arXiv:2508.10104 [cs.CV] Susan Sontag

    DINOv3. arXiv:2508.10104 [cs.CV] Susan Sontag. 2003.Regarding the Pain of Others. Farrar, Straus and Giroux. Katherine R. Storrs, Tim C. Kietzmann, Alexander Walther, Johannes Mehrer, and Nikolaus Kriegeskorte

    Pith/arXiv arXiv 2003

  35. [35]

    Francisco J Varela

    Diverse deep neural networks all predict human inferior temporal cortex well, after training and fitting.Journal of Cognitive Neuroscience33, 10 (2021), 2044–2064. Francisco J Varela

    2021

  36. [36]

    Tongzhou Wang and Phillip Isola

    Neurophenomenology: A methodological remedy for the hard problem.Journal of Consciousness Studies3, 4 (1996), 330–349. Tongzhou Wang and Phillip Isola

    1996

  37. [37]

    Daniel LK Yamins and James J DiCarlo

    The selective impairment of semantic memory.Quarterly Journal of Experimental Psychology 27, 4 (1975), 635–657. Daniel LK Yamins and James J DiCarlo

    1975

  38. [38]

    Using goal-driven deep learning models to understand sensory cortex.Nature Neuroscience19, 3 (2016), 356–365. Proc. ACM Comput. Graph. Interact. Tech., Vol. 9, No. 3, Article

    2016

  39. [39]

    Joanna Zylinska

    Dementia as a cultural metaphor.The Gerontologist54, 2 (2013), 258–267. Joanna Zylinska. 2020.AI Art: Machine Visions and Warped Dreams. Open Humanities Press. Proc. ACM Comput. Graph. Interact. Tech., Vol. 9, No. 3, Article

    2013

  40. [40]

    Publication date: July 2026

    2026