Pith. sign in

REVIEW 4 major objections 5 minor 105 references

The Generalist Brain Module: Module Repetition in Neural Networks in Light of the Minicolumn Hypothesis

T0 review · 4 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read This review argues that repeating one shared neural module across a network yields a generalist module with collective-intelligence properties such as robustness, adaptability, and generalization.

desk verdict An honest, useful review that frames parameter-shared repeated modules as a biologically inspired 'generalist module'; worth refereeing once the placeholder figure and causal-attribution overreach are fixed. read the letter →

arxiv 2507.12473 v1 pith:WRS2GXSA submitted 2025-07-01 q-bio.NC cs.LGcs.NE

Mia-Katrin Kvalsund , Mikkel Elle Lepper{o}d This is my paper
classification q-bio.NCcs.LGcs.NE
keywords minicolumnhypothesisthousandbrainstheoryneuralmodulerepetitionparametersharinggeneralistcollectiveintelligencemodularrobotsscalability
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This review argues that repeating the same neural module throughout a network, rather than building specialized parts, is a coherent architectural strategy with roots in the brain's cortical minicolumns. Drawing on collective-intelligence and swarm literature, the authors distinguish architectural repetition from parameter-shared repetition, and claim that parameter sharing is what consistently yields robustness, adaptability, and generalization. The central synthesized finding is that the repeated module tends to become a generalist: a simple, flexible function that can take on many roles in the ensemble. If right, this would make networks cheaper to train, easier to scale after deployment, and more tolerant of damage or novel inputs, potentially addressing energy and scalability problems in modern AI. The review is explicit that controlled causal evidence is still missing and that theoretical support lags behind the empirical hints.

What carries the argument

The central object is the cortical minicolumn, the proposal that the neocortex is built from repeated elementary columns of neurons, reimagined as thousands of voting models of objects. The review's analytical instrument is the distinction between architectural repetition, which reuses structure with separate parameters, and parameter-shared module repetition, where one functional unit is copied across the network. Parameter sharing reduces the total parameter count from $N \times M$ to $M$ and forces each copy to work with clones of itself in many contexts, which is what drives convergence to a generalist and produces the swarm-like benefits.

What would settle it

A matched-capacity controlled comparison would settle it: train a shared-module ensemble and an independent-parameter ensemble with the same connectivity and optimizer on identical robot-morphology tasks, and if the independent ensemble shows equal or better robustness, scalability, and zero-shot generalization, the central causal claim is falsified.

Watch

Extended reading notes

Core claim

The paper's thesis is that the neocortical minicolumn, an elementary repeated module, provides a workable template for AI architectures built from repetition. It surveys historical and current work and concludes that when the same functional module is reused with shared parameters, ensembles display collective-intelligence properties: robustness to damage, adaptability through role switching, zero-shot generalization to new conditions, and scalability to larger ensembles. Across the reviewed systems, the repeated module tends to converge toward a generalist rather than an expert: a simple, flexible problem solver that handles many positions, inputs, and bodies. The authors frame this as a synthesis and a research direction, noting that the benefits have been observed piecemeal and that theoretical explanations are lacking.

Load-bearing premise

The load-bearing premise, acknowledged as unproven in the opening of Section 5, is that the observed benefits are meaningfully caused by module repetition itself, not by fewer parameters or easier optimization.

Editorial extensions

If this is right

  • A shared-module network can be trained at small scale and enlarged after deployment without adding parameters, cutting training cost and energy use.
  • Deleting or damaging modules need not delete function, because remaining copies can shift roles, giving fault tolerance.
  • A policy trained on several bodies or perspectives can transfer zero-shot to new, out-of-distribution bodies.
  • Permutation-invariant integration lets identical modules swap roles when their inputs are permuted, preserving ensemble behavior.
  • Shared parameters constrain fine-tuning, because changing behavior for one role perturbs all roles, the so-called debugging problem.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the causal link is later established, the same architecture could be tested as a basis for continual learning, where roles are reassigned as tasks arrive without retraining the whole network.
  • The perspective mechanism suggests a quantitative prediction: generalization should improve as the number of input subspaces per task grows and should collapse when all modules receive identical input, a testable dose-response curve.
  • The review's caveat implies that some reported gains may be explained by reduced parameter count or by evolutionary search dynamics, so future work should separate those factors.
  • A natural extension is to use the minicolumn lens to design heterogeneous modular systems that start from a shared module and then specialize locally through Hebbian or local learning, as some reviewed works already begin to do.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. This review paper argues that neural module repetition, and specifically parameter-shared module repetition, is a brain-inspired architectural principle aligned with the minicolumn hypothesis, and that such repetition yields collective-intelligence properties including robustness, adaptability, generalization, scalability, and energy efficiency. The authors distinguish architectural repetition from parameter-shared repetition, survey a broad set of methods (convolutional blocks, neural architecture search, modular robotics, neural cellular automata, and others), and propose that the repeated module converges to a 'generalist module' capable of flexibly taking on many roles. They also present a speculative causal-model hypothesis for why repetition should aid out-of-distribution generalization, and they explicitly note that theoretical results are lacking.

Significance. If the central thesis is established, the paper would point to an architectural strategy that is simultaneously more biologically grounded, more parameter-efficient, more scalable, and more robust than monolithic networks. The review is valuable as a synthesis: it connects minicolumn neuroscience, collective intelligence, and a scattered AI literature on module repetition, and it provides a useful taxonomy (architectural vs. parameter-shared repetition) with tables of methods and tasks. The authors are appropriately cautious in some places, explicitly listing open questions in Section 5.4 and labeling the causal-generalization link a 'testable hypothesis.' However, the evidence presented does not yet support the strong causal claims that repetition itself produces the observed benefits, so the significance is conditional on those claims being substantiated.

major comments (4)
  1. [Section 5, opening paragraph] The review states that 'The effects must be meaningfully caused by the module repetition,' but the cited studies do not provide controlled comparisons that isolate repetition from other design choices. For instance, in Tang and Ha (2021) and Pedersen and Risi (2022), permutation invariance is produced by the aggregation mechanism (attention and summation, respectively) rather than by parameter-shared repetition per se; in Pathak et al. (2019) and Huang et al. (2020), parameter sharing is combined with graph message passing and training across multiple bodies; and in the evolved modular-robot works (e.g., Kvalsund et al. 2022, Mertan and Cheney 2023), comparisons to centralized controllers differ simultaneously in parameter count, decentralization, and optimization trajectory. Since the generalist-module thesis is the central claim of the review, this unresolved causal gap is load-bearing, and Section 5 should either present matched-pair evidence or explicitly reclassify these results as correlational observations.
  2. [Section 3.2] The causal-model hypothesis is introduced as 'a testable hypothesis' but is then given evidentiary support by an appeal to the same literature whose causal attribution is in question: 'Empirically, the reviewed works often showcase generalization, suggesting that there could be truth to the above hypothesis.' This is a circular move: the generalization results are the very phenomena whose mechanistic origin the hypothesis is supposed to explain. The review should either propose a concrete experimental design (e.g., matched architectures with and without parameter sharing) that could falsify the hypothesis, or present it purely as an open theoretical conjecture without claiming empirical support.
  3. [Section 5.1.1] The role-specialization results in Pedersen and Risi (2022) and Tang and Ha (2021) are presented as evidence for the generalist-module view, but the review's own description attributes the role-switching behavior to 'the identical modules receiving different input and the integration of the module outputs being permutation invariant.' That means the integration mechanism, not the repetition of the module per se, is what enables permutation-invariant role reassignment. The review should not count these results as evidence that module repetition itself causes the benefit, unless a comparison is made to a non-repeated architecture with the same integration mechanism.
  4. [Section 5.2] The scalability claims rely on a small number of demonstrations without baselines: Christensen (2006) scales from 50 to 1200 modules and Kvalsund et al. (2024) from 225 to 676, but in both cases the systems also use decentralized, self-organizing control, so scalability could stem from the self-organizing algorithm rather than from parameter-shared module repetition. The review should also report more precisely what 'without significant loss' means and whether the zero-shot scaling was compared to a non-repeated baseline. Without such details, the scalability benefit remains an association rather than an established effect of repetition.
minor comments (5)
  1. [Figure 1] Figure 1 contains multiple 'Lorem ipsum' placeholder text blocks, which makes the figure unreadable and suggests the manuscript is not yet in a complete, submittable form; these placeholders must be replaced with the intended explanatory text.
  2. [Table 2] Several entries in Table 2 are misspelled: 'ImagNet' should be 'ImageNet', 'CIF AR-10' should be 'CIFAR-10', and 'Wikidetox' should be 'WikiDetox'.
  3. [Section 4.2] The sentence 'The blueprints which specific modules should slot into which specific part of the blueprint' is grammatically incomplete and should be rewritten; it appears to describe blueprint-module assignment but the current wording is unclear.
  4. [Section 3.1] The search-space argument in Figure 5 is illustrated with 'imagined' fitness landscapes, but the figure lacks axis labels or a description of what the collapsed 'Phenotype' axis represents; this makes the argument harder to follow.
  5. [Abstract] The abstract states that parameter-shared module repetition 'exhibits key CI properties such as robustness, adaptability, and generalization,' but Section 5 shows that many of these properties are untested or only anecdotally observed; the wording should be softened to 'is reported to exhibit' or 'may exhibit'.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: the review is a synthesis of external literature, self-citations are illustrative rather than load-bearing, and the causal-generalization claim is explicitly framed as a testable hypothesis.

full rationale

This is a narrative review rather than a derivation, so there is no equation chain whose output is its input. The closest thing to a definitional slippage is the abstract's gloss of the generalist module as 'capable of handling many roles in the ensemble' for a parameter-shared module that is by definition repeated across roles; however, the review's substantive claims (robustness, adaptability, out-of-distribution generalization) are supported by independent empirical work (Tang and Ha 2021; Pedersen and Risi 2022; Pathak et al. 2019; Huang et al. 2020; Mertan and Cheney 2023; and others), and Section 3.2 explicitly labels the causal-generalization connection a 'testable hypothesis' rather than a derived theorem. The authors' own works (Kvalsund et al. 2022, 2024) are cited as examples alongside external studies, and removing them would not collapse the synthesis, so the self-citations are not load-bearing. The acknowledged absence of controlled comparisons isolating module repetition from parameter count, optimization method, and integration design is a causal-evidence limitation, not a circularity. Score 2 reflects minor non-load-bearing self-citations, not any reduction of the review's thesis to its inputs.

Assumptions & free parameters 0 free parameters · 4 assumptions · 0 invented entities

The review rests on domain assumptions (the minicolumn hypothesis, the transferability of swarm properties to artificial modules) and on a novel causal-model conjecture introduced in §3.2. It introduces no fitted parameters or invented physical entities.

assumptions (4)
  • domain assumption The minicolumn hypothesis is a valid functional model of the neocortex, and columns can be viewed as repeated generalist modules.
    The entire analogy to AI depends on taking Mountcastle's minicolumn hypothesis and Hawkins' Thousand Brains Theory as functionally meaningful; the review acknowledges the debate (Horton and Adams 2005) but proceeds on this assumption. Introduced in Section 1.
  • domain assumption Properties of swarm intelligence (simplicity, flexibility, robustness, scalability, parallel execution) transfer to artificial systems of repeated neural modules.
    Section 3.4 argues that a system of repeated modules fits Dorigo and Birattari's definition of a swarm, and therefore inherits swarm properties. This is an analogy, not a proven theorem.
  • ad hoc to paper If each module shares one parameter set, receives a coherent input subspace, and contributes to global action, then the module learns an approximate causal model of the data-generating process.
    This hypothesis is introduced in Section 3.2 as an extension of Richens and Everitt (2024); it is not proven and is presented as a 'testable hypothesis'.
  • domain assumption The reviewed sample of works is representative, and the benefits observed in them are meaningfully caused by module repetition.
    Stated in Section 5's opening: 'The effects must be meaningfully caused by the module repetition.' This is assumed rather than demonstrated by controlled comparison.

how reviews work

0 comments
Cite this review

Pith. "Pith review of The Generalist Brain Module: Module Repetition in Neural Networks in Light of the Minicolumn Hypothesis." pith.science (2026). https://pith.science/paper/WRS2GXSA

@misc{pith2026250712473,
  author       = {Pith},
  title        = {Pith review of: The Generalist Brain Module: Module Repetition in Neural Networks in Light of the Minicolumn Hypothesis},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WRS2GXSA}},
  note         = {Machine review of arXiv:2507.12473}
}
read the original abstract

While modern AI continues to advance, the biological brain remains the pinnacle of neural networks in its robustness, adaptability, and efficiency. This review explores an AI architectural path inspired by the brain's structure, particularly the minicolumn hypothesis, which views the neocortex as a distributed system of repeated modules - a structure we connect to collective intelligence (CI). Despite existing work, there is a lack of comprehensive reviews connecting the cortical column to the architectures of repeated neural modules. This review aims to fill that gap by synthesizing historical, theoretical, and methodological perspectives on neural module repetition. We distinguish between architectural repetition - reusing structure - and parameter-shared module repetition, where the same functional unit is repeated across a network. The latter exhibits key CI properties such as robustness, adaptability, and generalization. Evidence suggests that the repeated module tends to converge toward a generalist module: simple, flexible problem solvers capable of handling many roles in the ensemble. This generalist tendency may offer solutions to longstanding challenges in modern AI: improved energy efficiency during training through simplicity and scalability, and robust embodied control via generalization. While empirical results suggest such systems can generalize to out-of-distribution problems, theoretical results are still lacking. Overall, architectures featuring module repetition remain an emerging and unexplored architectural strategy, with significant untapped potential for both efficiency, robustness, and adaptiveness. We believe that a system that adopts the benefits of CI, while adhering to architectural and functional principles of the minicolumns, could challenge the modern AI problems of scalability, energy consumption, and democratization.

Figures

Figures reproduced from arXiv: 2507.12473 by the authors.

Figure 1
Figure 1. Key theoretical and empirical insights on the generalist module: a) We conceptualize the neocortex as a system embodying collective intelligence through its distributed, homogeneous column architecture. This suggests that systems like it can be analyzed using the literature on swarm and collective intelligence. b) Each module in the collective is a generalist (right), one that can flexibly adopt different skillsets … view at source ↗
Figure 2
Figure 2. Cortical columns in the sensory cortex each receive a limited part of the senses. Across the neocortex, the outer part of the mammalian brain, there are small units of neurons called cortical columns. They correspond to only a small part of the body, such that the whole sensory cortex corresponds to the whole body. their new model, Monty, from the Thousand Brains Project (George and Hawkins, 2009; Clay et al., 2024)… view at source ↗
Figure 3
Figure 3. Examples of module repetition this review will concern. The boxes labeled M are modules, where col￾ors signify that a specific architecture is used. a) Network of parameter sharing modules: A network repeating one module, M1, in multiple positions. b) We also include when repetition of modules is not fully homogeneous. Here, a network repeats two modules, M1 and M2. c) Disconnected and parallel mod￾ules with paramet… view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Views on the brain’s modularity. a) The various authors described in this section fall somewhere on the spectrum between believing in specialized versus general and modular versus integrated cognition. b) Each box marked M represents a brain module, with the color sign…
Figure 5
Figure 5. Figure 5: An example of a neural network with 81 weights without neural module repetition (Top) and with (Bottom). The graphs are the imagined search spaces, with the possible networks (Phenotypes) on the bottom axis and the score of the possible networks on the vertical axis. I…
Figure 6
Figure 6. Figure 6: Examples of perspectives on tasks. a) In image classification, pixel neighborhoods can be seen as perspectives on the task. Each perspective-label pair represents a subtask. The parameter set P is encouraged to output such that the global behavior classifies the image.…
Figure 7
Figure 7. Figure 7: An example of a modular robot. The black and blue physical module is repeated six times to create one creature that can move across a 3D environment. Photo: Part of ”Examples of well-performing morphologies” by Kvalsund et al. (2022). Photo licensed under CC BY 4.0 net…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

105 extracted references · 66 canonical work pages

  1. [1]

    , author Maul, T

    author Amer, M. , author Maul, T. , year 2019 . title A review of modularization techniques in artificial neural networks . journal Artificial Intelligence Review volume 52 , pages 527--561

    2019

  2. [2]

    , author Martin, K

    author Amorim Da Costa, N.M.M. , author Martin, K. , year 2010 . title Whose cortical column would that be? journal Frontiers in neuroanatomy , pages 16

    2010

  3. [3]

    , year 2007

    author Anderson, M.L. , year 2007 . title The massive redeployment hypothesis and the functional topography of the brain . journal Philosophical Psychology volume 20 , pages 143--174

    2007

  4. [4]

    , year 2010

    author Anderson, M.L. , year 2010 . title Neural reuse: A fundamental organizational principle of the brain . journal Behavioral and brain sciences volume 33 , pages 245--266

    2010

  5. [5]

    , author Finlay, B.L

    author Anderson, M.L. , author Finlay, B.L. , year 2014 . title Allocating structure to function: the strong links between neuroplasticity and natural selection . journal Frontiers in human neuroscience volume 7 , pages 918

    2014

  6. [6]

    , author Donti, P

    author Bashir, N. , author Donti, P. , author Cuff, J. , author Sroka, S. , author Ilic, M. , author Sze, V. , author Delimitrou, C. , author Olivetti, E. , year 2024 . title The Climate and Sustainability Implications of Generative AI . journal An MIT Exploration of Generative AI note Https://mit-genai.pubpub.org/pub/8ulgrckc

    2024

  7. [7]

    , author Gebru, T

    author Bender, E.M. , author Gebru, T. , author McMillan-Major, A. , author Shmitchell, S. , year 2021 . title On the dangers of stochastic parrots: Can language models be too big? , in: booktitle Proceedings of the 2021 ACM conference on fairness, accountability, and transparency , pp. pages 610--623

    2021

  8. [8]

    , author Wang, J

    author Beni, G. , author Wang, J. , year 1993 . title Swarm intelligence in cellular robotic systems , in: booktitle Robots and biological systems: towards a new bionics? . publisher Springer , pp. pages 703--712

    1993

Show all 105 references
  1. [9]

    , year 2019

    author Berkeley, I.S. , year 2019 . title The curious case of connectionism . journal Open Philosophy volume 2 , pages 190--205

    2019

  2. [10]

    , author Dorigo, M

    author Bonabeau, E. , author Dorigo, M. , author Theraulaz, G. , author Theraulaz, G. , year 1999 . title Swarm intelligence: from natural to artificial systems . number 1 , publisher Oxford university press

    1999

  3. [11]

    , year 1986

    author Brooks, R.A. , year 1986 . title A robust layered control system for a mobile robot . journal IEEE journal on robotics and automation volume 2 , pages 14--23

    1986

  4. [12]

    , year 2014

    author Brooks, R.A. , year 2014 . title How to build complete creatures rather than isolated cognitive simulators , in: booktitle Architectures for intelligence . publisher Psychology Press , pp. pages 239--254

    2014

  5. [13]

    , author Kar, A.K

    author Chakraborty, A. , author Kar, A.K. , year 2017 . title Swarm intelligence: A review of algorithms . journal Nature-inspired computing and optimization , pages 475--494

    2017

  6. [14]

    , author Bongard, J

    author Cheney, N. , author Bongard, J. , author SunSpiral, V. , author Lipson, H. , year 2016 . title On the difficulty of co-optimizing morphology and control in evolved virtual creatures , in: booktitle Artificial life conference proceedings , organization MIT Press One Roge...

    2016

  7. [15]

    , author Shimohara, K

    author Cho, S.B. , author Shimohara, K. , year 1998 . title Evolutionary learning of modular neural networks with genetic programming . journal Applied Intelligence volume 9 , pages 191--200

    1998

  8. [16]

    , author Knoop, M

    author Chollet, F. , author Knoop, M. , author Kamradt, G. , author Landers, B. , year 2024 . title Arc prize 2024: Technical report . journal arXiv preprint arXiv:2412.04604

    2024 arXiv

  9. [17]

    , year 2006

    author Christensen, D.J. , year 2006 . title Evolution of shape-changing and self-repairing control for the ATRON self-reconfigurable robot , in: booktitle Proceedings 2006 IEEE International Conference on Robotics and Automation, 2006. ICRA 2006. , organization IEEE . pp. pag...

    2006

  10. [18]

    , author Leadholm, N

    author Clay, V. , author Leadholm, N. , author Hawkins, J. , year 2024 . title The thousand brains project: A new paradigm for sensorimotor intelligence . journal arXiv preprint arXiv:2412.18354

    2024 arXiv

  11. [19]

    , author Tooby, J

    author Cosmides, L. , author Tooby, J. , year 1994 . title Origins of domain specificity: The evolution of functional organization . journal Mapping the mind: Domain specificity in cognition and culture volume 853116

    1994

  12. [20]

    , author Meyer, J.A

    author Doncieux, S. , author Meyer, J.A. , year 2004 . title Evolving modular neural networks to solve challenging control problems , in: booktitle Proceedings of the Fourth International ICSC Symposium on engineering of intelligent systems (EIS 2004) , pp. pages 1--7

    2004

  13. [21]

    , author Birattari, M

    author Dorigo, M. , author Birattari, M. , year 2007 . title Swarm intelligence . journal Scholarpedia volume 2 , pages 1462 . :10.4249/scholarpedia.1462. note revision \#138640

    2007 doi

  14. [22]

    , year 2024

    author Egeland, J. , year 2024 . title Making sense of the modularity debate . journal New Ideas in Psychology volume 75 , pages 101108

    2024

  15. [23]

    , author Akaishi, R

    author Eisenreich, B.R. , author Akaishi, R. , author Hayden, B.Y. , year 2017 . title Control without controllers: toward a distributed neuroscience of executive control . journal Journal of cognitive neuroscience volume 29 , pages 1684--1698

    2017

  16. [24]

    , author Iacca, G

    author Ferigo, A. , author Iacca, G. , author Medvet, E. , author Nadizar, G. , year 2025 . title Totipotent neural controllers for modular soft robots: Achieving specialization in body--brain co-evolution through hebbian learning . journal Neurocomputing volume 614 , pages 128811

    2025

  17. [25]

    , year 1983

    author Fodor, J.A. , year 1983 . title The modularity of mind . publisher MIT press

    1983

  18. [26]

    , year 2000

    author Fodor, J.A. , year 2000 . title The mind doesn't work that way: The scope and limits of computational psychology . publisher MIT press

    2000

  19. [27]

    , author Ploeger, A

    author Frankenhuis, W.E. , author Ploeger, A. , year 2007 . title Evolutionary psychology versus fodor: Arguments for and against the massive modularity hypothesis . journal Philosophical Psychology volume 20 , pages 687--710

    2007

  20. [28]

    , author Ramstead, M.J

    author Friston, K.J. , author Ramstead, M.J. , author Kiefer, A.B. , author Tschantz, A. , author Buckley, C.L. , author Albarracin, M. , author Pitliya, R.J. , author Heins, C. , author Klein, B. , author Millidge, B. , et al., year 2024 . title Designing ecosystems of intell...

    2024

  21. [29]

    , year 1907

    author Galton, F. , year 1907 . title Vox populi (the wisdom of crowds) . journal Nature volume 75 , pages 450--451

    1907

  22. [30]

    , author George, A.H

    author George, A.S. , author George, A.H. , author Martin, A.G. , year 2023 . title The environmental impact of ai: A case study of water consumption by chat gpt . journal Partners Universal International Innovation Journal volume 1 , pages 97--104

    2023

  23. [31]

    , author Hawkins, J

    author George, D. , author Hawkins, J. , year 2009 . title Towards a mathematical theory of cortical micro-circuits . journal PLoS computational biology volume 5 , pages e1000532

    2009

  24. [32]

    , author Khosravi, H

    author Gholamalinezhad, H. , author Khosravi, H. , year 2020 . title Pooling methods in deep neural networks, a review . journal arXiv preprint arXiv:2009.07485

    2020 arXiv

  25. [33]

    , author Bengio, Y

    author Goodfellow, I. , author Bengio, Y. , author Courville, A. , author Bengio, Y. , year 2016 . title Deep learning . volume volume 1 . publisher MIT press Cambridge . note http://www.deeplearningbook.org

    2016

  26. [34]

    , author Bengio, Y

    author Goyal, A. , author Bengio, Y. , year 2022 . title Inductive biases for deep learning of higher-level cognition . journal Proceedings of the Royal Society A volume 478 , pages 20210068

    2022

  27. [35]

    , author Livi, L

    author Grattarola, D. , author Livi, L. , author Alippi, C. , year 2021 . title Learning graph cellular automata . journal Advances in Neural Information Processing Systems volume 34 , pages 20983--20994

    2021

  28. [36]

    , year 1995

    author Greenblatt, S.H. , year 1995 . title Phrenology in the science and culture of the 19th century . journal Neurosurgery volume 37 , pages 790--805

    1995

  29. [37]

    , author Tang, Y

    author Ha, D. , author Tang, Y. , year 2022 . title Collective intelligence for deep learning: A survey of recent developments . journal Collective Intelligence volume 1 , pages 26339137221114874

    2022

  30. [38]

    , year 2021

    author Hawkins, J. , year 2021 . title A thousand brains: A new theory of intelligence . publisher Hachette UK

    2021

  31. [39]

    , author Lewis, M

    author Hawkins, J. , author Lewis, M. , author Klukas, M. , author Purdy, S. , author Ahmad, S. , year 2019 . title A framework for intelligence and cortical function based on grid cells in the neocortex . journal Frontiers in neural circuits , pages 121

    2019

  32. [40]

    , author Zhang, X

    author He, K. , author Zhang, X. , author Ren, S. , author Sun, J. , year 2016 . title Deep residual learning for image recognition , in: booktitle Proceedings of the IEEE conference on computer vision and pattern recognition , pp. pages 770--778

    2016

  33. [41]

    , year 2021

    author Hiesinger, P.R. , year 2021 . title The self-assembling brain . publisher Princeton university press

    2021

  34. [42]

    , author Ahmad, S

    author Hole, K.J. , author Ahmad, S. , year 2021 . title A thousand brains: toward biologically constrained ai . journal SN Applied Sciences volume 3 , pages 743

    2021

  35. [43]

    , author Lipson, H

    author Hornby, G.S. , author Lipson, H. , author Pollack, J.B. , year 2003 . title Generative representations for the automated design of modular physical robots . journal IEEE transactions on Robotics and Automation volume 19 , pages 703--719

    2003

  36. [44]

    , author Adams, D.L

    author Horton, J.C. , author Adams, D.L. , year 2005 . title The cortical column: a structure without a function . journal Philosophical Transactions of the Royal Society B: Biological Sciences volume 360 , pages 837--862

    2005

  37. [45]

    , author Mordatch, I

    author Huang, W. , author Mordatch, I. , author Pathak, D. , year 2020 . title One policy to control them all: Shared modular policies for agent-agnostic control , in: booktitle International Conference on Machine Learning , organization PMLR . pp. pages 4455--4464

    2020

  38. [46]

    , year 1846

    author Hytche, E. , year 1846 . title Anatomy, physiology, and phrenology. journal The Lancet volume 47 , pages 12--13

  39. [47]

    , author Kong, S.G

    author Jiang, W. , author Kong, S.G. , year 2007 . title Block-based neural networks for personalized ecg signal classification . journal IEEE Transactions on Neural Networks volume 18 , pages 1750--1761

    2007

  40. [48]

    , author Kong, S.G

    author Jiang, W. , author Kong, S.G. , author Peterson, G.D. , year 2005 . title Ecg signal classification using block-based neural networks , in: booktitle Proceedings. 2005 IEEE International Joint Conference on Neural Networks, 2005. , organization IEEE . pp. pages 326--331

    2005

  41. [49]

    , author Alfaro-Almagro, F

    author Jones, O.P. , author Alfaro-Almagro, F. , author Jbabdi, S. , year 2018 . title An empirical, 21st century evaluation of phrenology . journal Cortex volume 106 , pages 26--35

    2018

  42. [50]

    , author Snorek, M

    author Koutník, J. , author Snorek, M. , year 2004 . title Single categorizing and learning module for temporal sequences , in: booktitle 2004 IEEE International Joint Conference on Neural Networks (IEEE Cat. No. 04CH37541) , organization IEEE . pp. pages 2977--2982

    2004

  43. [51]

    , author Ruxton, G.D

    author Krause, J. , author Ruxton, G.D. , author Krause, S. , year 2010 . title Swarm intelligence in animals and humans . journal Trends in ecology & evolution volume 25 , pages 28--34

    2010

  44. [52]

    , author Glette, K

    author Kvalsund, M.K. , author Glette, K. , author Veenstra, F. , year 2022 . title Centralized and decentralized control in modular robots and their effect on morphology , in: booktitle Artificial Life Conference Proceedings 34 , organization MIT Press One Rogers Street, Camb...

    2022

  45. [53]

    , author Pontes-Filho, S

    author Kvalsund, M.K. , author Pontes-Filho, S. , author Glette, K. , author Ellefsen, K.O. , author Lepper d, M.E. , year 2024 . title Sensor movement drives emergent attention and scalability in active neural cellular automata . journal bioRxiv , pages 2024--12

    2024

  46. [54]

    , author Maire, M

    author Larsson, G. , author Maire, M. , author Shakhnarovich, G. , year 2016 . title Fractalnet: Ultra-deep neural networks without residuals . journal arXiv preprint arXiv:1605.07648

    2016 arXiv

  47. [55]

    , author Boser, B

    author LeCun, Y. , author Boser, B. , author Denker, J.S. , author Henderson, D. , author Howard, R.E. , author Hubbard, W. , author Jackel, L.D. , year 1989 . title Backpropagation applied to handwritten zip code recognition . journal Neural computation volume 1 , pages 541--551

    1989

  48. [56]

    , author Meyerson, E

    author Liang, J. , author Meyerson, E. , author Hodjat, B. , author Fink, D. , author Mutch, K. , author Miikkulainen, R. , year 2019 . title Evolutionary neural automl for deep learning , in: booktitle Proceedings of the Genetic and Evolutionary Computation Conference , pp. p...

    2019

  49. [57]

    , et al., year 2007

    author Lipson, H. , et al., year 2007 . title Principles of modularity, regularity, and hierarchy for scalable systems . journal Journal of Biological Physics and Chemistry volume 7 , pages 125

    2007

  50. [58]

    , author Simonyan, K

    author Liu, H. , author Simonyan, K. , author Vinyals, O. , author Fernando, C. , author Kavukcuoglu, K. , year 2017 . title Hierarchical representations for efficient architecture search . journal arXiv preprint arXiv:1711.00436

    2017 arXiv

  51. [59]

    , author Zhang, S.J

    author Long, X. , author Zhang, S.J. , year 2021 . title A novel somatosensory spatial navigation system outside the hippocampal formation . journal Cell research volume 31 , pages 649--663

    2021

  52. [60]

    , year 2023

    author Lu, Z. , year 2023 . title A theory of multimodal learning . journal Advances in Neural Information Processing Systems volume 36 , pages 57244--57255

    2023

  53. [61]

    , year 2023

    author McCaffrey, J.B. , year 2023 . title Evolving concepts of functional localization . journal Philosophy Compass volume 18 , pages e12914

    2023

  54. [62]

    , author Levin, M

    author McMillen, P. , author Levin, M. , year 2024 . title Collective intelligence: A unifying concept for integrating biology across scales and substrates . journal Communications Biology volume 7 , pages 378

    2024

  55. [63]

    , author Cheney, N

    author Mertan, A. , author Cheney, N. , year 2023 . title Modular controllers facilitate the co-optimization of morphology and control in soft robots , in: booktitle Proceedings of the Genetic and Evolutionary Computation Conference , pp. pages 174--183

    2023

  56. [64]

    , author Liang, J

    author Miikkulainen, R. , author Liang, J. , author Meyerson, E. , author Rawal, A. , author Fink, D. , author Francon, O. , author Raju, B. , author Shahrzad, H. , author Navruzyan, A. , author Duffy, N. , author Hodjat, B. , year 2017 . title Evolving deep neural networks . ...

    2017 arXiv

  57. [65]

    , year 1988

    author Minsky, M. , year 1988 . title Society of mind . publisher Simon and Schuster

    1988

  58. [66]

    , author Kong, S.G

    author Moon, S.W. , author Kong, S.G. , year 2001 . title Block-based neural networks . journal IEEE Transactions on Neural Networks volume 12 , pages 307--317

    2001

  59. [67]

    , author Randazzo, E

    author Mordvintsev, A. , author Randazzo, E. , author Niklasson, E. , author Levin, M. , year 2020 . title Growing neural cellular automata . journal Distill volume 5 , pages e23

    2020

  60. [68]

    , year 1978

    author Mountcastle, V. , year 1978 . title An organizing principle for cerebral function: the unit module and the distributed system . journal The mindful brain

    1978

  61. [69]

    , year 1997

    author Mountcastle, V. , year 1997 . title The columnar organization of the neocortex. journal Brain: a journal of neurology volume 120 , pages 701--722

    1997

  62. [70]

    , author Doncieux, S

    author Mouret, J.B. , author Doncieux, S. , year 2008 . title Mennag: a modular, regular and hierarchical encoding for neural-networks based on attribute grammars . journal Evolutionary Intelligence volume 1 , pages 187--207

    2008

  63. [71]

    , author Doncieux, S

    author Mouret, J.B. , author Doncieux, S. , author Meyer, J.A. , year 2006 . title Incremental evolution of target-following neuro-controllers for flapping-wing animats , in: booktitle International Conference on Simulation of Adaptive Behavior , organization Springer . pp. pa...

    2006

  64. [72]

    , author Nazari, M

    author Mousavi, H.K. , author Nazari, M. , author Tak \'a c , M. , author Motee, N. , year 2019 . title Multi-agent image classification via reinforcement learning , in: booktitle 2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , organization IE...

    2019

  65. [73]

    , author Phaf, R.H

    author Murre, J.M. , author Phaf, R.H. , author Wolters, G. , year 1992 . title Calm: Categorizing and learning module . journal Neural Networks volume 5 , pages 55--82

    1992

  66. [74]

    From Cells to Societies: Collective Learning across Scales

    author Nadizar, G. , author Medvet, E. , author Nichele, S. , author Pontes-Filho, S. , et al., year 2022 . title Collective control of modular soft robots via embodied spiking neural cellular automata , in: booktitle Proceedings of the workshop on" From Cells to Societies: Co...

    2022

  67. [75]

    , author Lu, C

    author Pathak, D. , author Lu, C. , author Darrell, T. , author Isola, P. , author Efros, A.A. , year 2019 . title Learning to control self-assembling morphologies: a study of generalization via modularity . journal Advances in Neural Information Processing Systems volume 32

    2019

  68. [76]

    , author Plantec, E

    author Pedersen, J. , author Plantec, E. , author Nisioti, E. , author Montero, M. , author Risi, S. , year 2024 . title Structurally flexible neural networks: Evolving the building blocks for general agents , in: booktitle Proceedings of the Genetic and Evolutionary Computati...

    2024

  69. [77]

    , author Risi, S

    author Pedersen, J.W. , author Risi, S. , year 2022 . title Minimal neural network models for permutation invariant agents , in: booktitle Proceedings of the Genetic and Evolutionary Computation Conference , pp. pages 130--138

    2022

  70. [78]

    , year 2005

    author Pinker, S. , year 2005 . title So how does the mind work? journal Mind & language volume 20 , pages 1--24

    2005

  71. [79]

    , author Mordvintsev, A

    author Randazzo, E. , author Mordvintsev, A. , author Niklasson, E. , author Levin, M. , author Greydanus, S. , year 2020 . title Self-classifying mnist digits . journal Distill volume 5 , pages e00027--002

    2020

  72. [80]

    , author Stanley, K.O

    author Reisinger, J. , author Stanley, K.O. , author Miikkulainen, R. , year 2004 . title Evolving reusable neural modules , in: booktitle Genetic and Evolutionary Computation Conference , organization Springer . pp. pages 69--81

    2004

  73. [81]

    , author Everitt, T

    author Richens, J. , author Everitt, T. , year 2024 . title Robust agents learn causal world models . journal arXiv preprint arXiv:2402.10877

    2024 arXiv

  74. [82]

    , year 2021

    author Risi, S. , year 2021 . title The future of artificial intelligence is self-organizing and self-assembling . journal sebastianrisi.com https://sebastianrisi.com/self_assembling_ai

    2021

  75. [83]

    , author Andreas, J

    author Ruis, L. , author Andreas, J. , author Baroni, M. , author Bouchacourt, D. , author Lake, B.M. , year 2020 . title A benchmark for systematic generalization in grounded language understanding . journal Advances in neural information processing systems volume 33 , pages ...

    2020

  76. [84]

    , author McClelland, J.L

    author Rumelhart, D.E. , author McClelland, J.L. , author Group, P.R. , et al., year 1988 . title Parallel distributed processing . volume volume 1 . publisher IEEE New York

    1988

  77. [85]

    , author Ling, S.H

    author San, P.P. , author Ling, S.H. , author Nguyen, H. , et al., year 2013 . title Evolvable rough-block-based neural network and its biomedical application to hypoglycemia detection system . journal IEEE transactions on cybernetics volume 44 , pages 1338--1349

    2013

  78. [86]

    , author Dodge, J

    author Schwartz, R. , author Dodge, J. , author Smith, N.A. , author Etzioni, O. , year 2020 . title Green ai . journal Communications of the ACM volume 63 , pages 54--63

    2020

  79. [87]

    , author Paik, J

    author Seo, J. , author Paik, J. , author Yim, M. , year 2019 . title Modular reconfigurable robotics . journal Annual Review of Control, Robotics, and Autonomous Systems volume 2 , pages 63--88

    2019

  80. [88]

    , author Mashoufi, B

    author Shadmand, S. , author Mashoufi, B. , year 2016 . title A new personalized ecg signal classification algorithm using block-based neural network and particle swarm optimization . journal Biomedical Signal Processing and Control volume 25 , pages 12--23

    2016

  81. [89]

    , author Greff, K

    author Srivastava, R.K. , author Greff, K. , author Schmidhuber, J. , year 2015 . title Highway networks . journal arXiv preprint arXiv:1505.00387

    2015 arXiv

  82. [90]

    , author Shen, W.M

    author Stoy, K. , author Shen, W.M. , author Will, P.M. , year 2002 . title Using role-based control to produce locomotion in chain-type self-reconfigurable robots . journal IEEE/ASME transactions on mechatronics volume 7 , pages 410--417

    2002

  83. [91]

    , author Ganesh, A

    author Strubell, E. , author Ganesh, A. , author McCallum, A. , year 2019 . title Energy and policy considerations for deep learning in nlp . journal arXiv preprint arXiv:1906.02243

    2019 arXiv

  84. [92]

    , author Pratt, S.C

    author Sumpter, D.J. , author Pratt, S.C. , year 2009 . title Quorum responses and consensus decision making . journal Philosophical Transactions of the Royal Society B: Biological Sciences volume 364 , pages 743--753

    2009

  85. [93]

    , author Pennartz, C.M

    author Suzuki, M. , author Pennartz, C.M. , author Aru, J. , year 2023 . title How deep is the brain? the shallow brain hypothesis . journal Nature Reviews Neuroscience volume 24 , pages 778--791

    2023

  86. [94]

    , author Liu, W

    author Szegedy, C. , author Liu, W. , author Jia, Y. , author Sermanet, P. , author Reed, S. , author Anguelov, D. , author Erhan, D. , author Vanhoucke, V. , author Rabinovich, A. , year 2015 . title Going deeper with convolutions , in: booktitle Proceedings of the IEEE confe...

    2015

  87. [95]

    , author Ha, D

    author Tang, Y. , author Ha, D. , year 2021 . title The sensory neuron as a transformer: Permutation-invariant neural networks for reinforcement learning . journal Advances in Neural Information Processing Systems volume 34 , pages 22574--22587

    2021

  88. [96]

    , author Greenewald, K

    author Thompson, N.C. , author Greenewald, K. , author Lee, K. , author Manso, G.F. , et al., year 2020 . title The computational limits of deep learning . journal arXiv preprint arXiv:2007.05558 volume 10

    2020 arXiv

  89. [97]

    , author Erez, T

    author Todorov, E. , author Erez, T. , author Tassa, Y. , year 2012 . title Mujoco: A physics engine for model-based control , in: booktitle 2012 IEEE/RSJ International Conference on Intelligent Robots and Systems , organization IEEE . pp. pages 5026--5033 . :10.1109/IROS.2012.6386109

    2012

  90. [98]

    , author Jiang, F

    author Tran, Q.A. , author Jiang, F. , author Hu, J. , year 2012 . title A real-time netflow-based intrusion detection system with improved bbnn and high-frequency field programmable gate arrays , in: booktitle 2012 IEEE 11th International Conference on Trust, Security and Pri...

    2012

  91. [99]

    , author Pontes-Filho, S

    author Variengien, A. , author Pontes-Filho, S. , author Glover, T.E. , author Nichele, S. , year 2021 . title Towards self-organized control: Using neural cellular automata to robustly control a cart-pole agent . journal Innovations in Machine Intelligence (IMI) volume 1 , pa...

    2021 doi

  92. [100]

    , author Glette, K

    author Veenstra, F. , author Glette, K. , year 2020 . title How different encodings affect performance and diversification when evolving the morphology and control of 2d virtual creatures , in: booktitle Artificial Life Conference Proceedings 32 , organization MIT Press One Ro...

    2020

  93. [101]

    , author Escola, S

    author Zador, A. , author Escola, S. , author Richards, B. , author \"O lveczky, B. , author Bengio, Y. , author Boahen, K. , author Botvinick, M. , author Chklovskii, D. , author Churchland, A. , author Clopath, C. , et al., year 2022 . title Toward next-generation artificial...

    2022 arXiv

  94. [102]

    , year 2017

    author Zerilli, J. , year 2017 . title Against the “system” module . journal Philosophical Psychology volume 30 , pages 235--250

    2017

  95. [103]

    , year 2019

    author Zerilli, J. , year 2019 . title Neural reuse and the modularity of mind: where to next for modularity? journal Biological Theory volume 14 , pages 1--20

    2019

  96. [104]

    , author Yang, Z

    author Zhong, Z. , author Yang, Z. , author Deng, B. , author Yan, J. , author Wu, W. , author Shao, J. , author Liu, C.L. , year 2020 . title Blockqnn: Efficient block-wise neural network architecture generation . journal IEEE transactions on pattern analysis and machine inte...

    2020

  97. [105]

    , author Le, Q.V

    author Zoph, B. , author Le, Q.V. , year 2016 . title Neural architecture search with reinforcement learning . journal arXiv preprint arXiv:1611.01578

    2016 arXiv

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.