pith. sign in

arxiv: 2605.16321 · v1 · pith:K5RDLY45new · submitted 2026-05-05 · 💻 cs.LG

Language Game: Talking to Non-Human Systems

Yanbo Zhang , Michael Levin This is my paper

Pith reviewed 2026-05-21 00:45 UTC · model grok-4.3

classification 💻 cs.LG
keywords language gamesgene regulatory networksreinforcement learningnon-neural intelligencedynamical systemsdialogue frameworksreservoir dynamics
0
0 comments X

The pith

Non-neural systems such as gene regulatory networks can engage in fluent dialogue through language games without any changes to their internal parameters.

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

The paper tries to establish a method for direct communication with diverse non-neural systems by treating interaction as a Wittgenstein-style language game. The system's dynamics stay frozen and function as the nonlinear core of a reinforcement learning policy, with only linear input and output interfaces trained via rewards so that states acquire meaning through use. This lets the system respond in its own behavior rather than through a proxy model. A sympathetic reader would care because it opens a route to accessing computation and decision-making in substrates like microbial consortia or fungi on their own terms. If the claim holds, behaviors from agents of different origins become comparable as different routes to the same game reward.

Core claim

By freezing the internal dynamics of systems such as gene regulatory networks as the nonlinear core of a reinforcement-learning policy and training only linear input and output interfaces, the framework yields fluent dialogue in which the system replies through its own behavior; different architectures converge on similar responses when pursuing the same reward, and specific GRN properties act as an inductive bias that makes some systems easier or harder to converse with.

What carries the argument

The language game that serves as a lingua franca across architectures, with the system's frozen dynamics acting as the nonlinear core of an RL policy whose states and actions acquire meaning through reward.

If this is right

  • Well-trained agents from disparate origins converge on similar behaviors when they optimize the same game reward.
  • Specific properties of gene regulatory networks determine how readily a system can participate in dialogue.
  • Fluent interaction becomes possible with any dynamical system whose dynamics can serve as an RL core, without parameter alteration.
  • Responses from otherwise irreconcilable representations can be read uniformly as pursuit of the shared game reward.

Where Pith is reading between the lines

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

  • The same embedding could be applied to other biological dynamical systems such as fungal networks to test for responsive decision-making.
  • Performance differences across GRNs might serve as a practical assay for their computational or memory properties.
  • Shared game semantics could provide a common metric for comparing goal-directed behavior between biological and artificial agents.
  • Scaling the linear interfaces to handle longer interaction sequences would test whether the approach remains stable without internal changes.

Load-bearing premise

A language model can reliably route a human prompt to a matching game and design an environmental state in which the system's desired action becomes the rational response to the game's reward.

What would settle it

If training the linear interfaces on a well-characterized gene regulatory network consistently fails to produce coherent, reward-aligned responses to designed prompts, the central claim would be falsified.

Figures

Figures reproduced from arXiv: 2605.16321 by Michael Levin, Yanbo Zhang.

Figure 1
Figure 1. Figure 1: (A) Two equivalent representations of the same Tic-Tac-Toe game state. A human perceives the game spa￾tially (three in a row wins), while an alien perceives it numerically (three numbers from a magic square summing to 15 wins). Because both descriptions encode identical rules, neither player need comprehend the other’s representation; the game itself acts as a lingua franca between incompatible cognitive a… view at source ↗
Figure 2
Figure 2. Figure 2: The Language Game framework. (A) The policy πf is a composition of a linear encoder E, the fixed sys￾tem dynamics f (evaluated as an instantaneous gradient), and a linear decoder D. The agent is trained on RL tasks to align the system’s dynamics with the tasks’ goals. This alignment allows translation between human prompts pH and environment states s, and from system actions aS back to human language pS. (… view at source ↗
Figure 3
Figure 3. Figure 3: Pairwise policy similarity (cosine) across 17 reservoir architectures on six representative environments span [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Training rewards across six representative RL tasks for different GRN reservoirs. [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: GRN properties as inductive biases for reinforcement learning. [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Per-cell 2 × 2 contingency-table heatmap. Each cell corresponds to one (GRN property, environment) pair; color encodes the signed − log10 of the raw two-sided Fisher-exact p-value, with red indicating that the property is over-represented in the top-seven (good) half of GRNs on that environment and blue that it is over-represented in the bottom-seven (bad) half. Small dots mark cells with nominal p < 0.10;… view at source ↗
Figure 7
Figure 7. Figure 7: Training reward curves for all 16 RL environments with 17 reservoir architectures. Dashed lines indicate [PITH_FULL_IMAGE:figures/full_fig_p021_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Policy similarity across all 17 reservoir architectures on all 16 RL environments. Each heatmap shows the [PITH_FULL_IMAGE:figures/full_fig_p026_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: System prompt for the router agent, which selects the most semantically appropriate RL environment for a [PITH_FULL_IMAGE:figures/full_fig_p028_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: System prompt for G(pH), which infers the goal action from human natural language. Here {env name} is the name of the selected environment, and {env desc} is a detailed description of the environment. Given env_name RL environment and a goal action, you design a environment state that lead a rational agent to ,→ take the goal action. # Environment Description: env_desc # Requirement and Instruction You wi… view at source ↗
Figure 11
Figure 11. Figure 11: System prompt for SLLM, which designs the environment state that leads a rational agent to take the goal action. Here {env name} is the name of the selected environment, and {env desc} is a detailed description of the environment. 28 [PITH_FULL_IMAGE:figures/full_fig_p028_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: System prompt for TS→H, which translates the system’s action into a natural language reply. Here {env name} is the name of the selected environment, and {env desc} is a detailed description of the environ￾ment. 29 [PITH_FULL_IMAGE:figures/full_fig_p029_12.png] view at source ↗
read the original abstract

Language carries thought and coordination among humans but rarely reaches further along the spectrum of diverse intelligence. Yet non-neural systems -- from gene regulatory networks and microbial consortia to fungi -- are increasingly recognized as substrates of computation, decision-making and memory, making dialogue with non-human intelligence newly conceivable. Today such dialogue is attempted only by proxy: a large language model speaks on the system's behalf, so any intelligence on display originates from the model while the system itself remains silent. Here we ask whether the system can speak in its own voice. Following Wittgenstein, who located meaning in use, we treat communication as a game played with the system. Its internal dynamics are frozen as the nonlinear core of a reinforcement-learning policy, with only linear input and output interfaces trained. Through use and reward, the system's states and responses acquire meaning within the game, so playing becomes speaking. Because different architectures playing the same game optimize the same reward, their behaviors can all be read as pursuit of that reward; the game serves as a lingua franca across otherwise irreconcilable representations. Given a human prompt, a language model routes it to the game whose semantics best match it and designs an environmental state for which the desired action is the rational response, letting the system reply through its own behavior. Applied across diverse gene regulatory networks and reinforcement-learning tasks, the framework yields fluent dialogue without altering any system parameter, shows that well-trained agents of disparate origin converge on similar behavior, and reveals that specific GRN properties make a system easier or harder to talk with -- an inductive bias of the reservoir itself. Our framework opens a new route to conversing with any dynamical system on its own terms.

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

2 major / 2 minor

Summary. The manuscript proposes a framework for conversing with non-human dynamical systems (e.g., gene regulatory networks) by treating communication as a Wittgensteinian language game. The target system's internal dynamics are frozen as the nonlinear core of an RL policy while only linear input/output interfaces are trained; an external LLM routes human prompts to pre-defined games and constructs environmental states such that the system's native trajectory realizes the desired reward-maximizing behavior. The authors claim that this yields fluent dialogue without parameter changes to the system, produces behavioral convergence across disparate architectures playing the same game, and exposes inductive biases of the reservoir itself.

Significance. If the central claims are empirically substantiated, the approach would offer a principled route to interpreting and interacting with non-neural computational substrates on their own terms, using games as a lingua franca. The emphasis on meaning arising from use and the separation of frozen nonlinear dynamics from trainable linear interfaces are conceptually attractive and could stimulate work at the intersection of RL, dynamical systems, and philosophy of language.

major comments (2)
  1. [Abstract] Abstract: the assertion that the framework 'yields fluent dialogue without altering any system parameter' and 'shows that well-trained agents of disparate origin converge on similar behavior' is presented without any quantitative metrics, success rates, example trajectories, or statistical comparisons. Because these outcomes are load-bearing for the central claim that the non-human system speaks in its own voice, the absence of supporting data or analysis undermines evaluation of the result.
  2. [Framework description] Framework description (abstract and pipeline): the LLM is described as routing the prompt to a game whose reward semantics match and then synthesizing an environmental state 'for which the desired action is the rational response.' This construction appears to pre-encode the intended semantics in the choice of game and state, so that the observed trajectory is an elicited rather than open-ended response. A concrete ablation or control experiment isolating the system's contribution from the LLM's preparatory work is required to support the claim that meaning emerges solely from the system's inductive bias and the game.
minor comments (2)
  1. [Abstract] The abstract introduces several terms ('fluent dialogue', 'inductive bias of the reservoir', 'lingua franca') without operational definitions or references to prior usage in the RL or dynamical-systems literature; a short definitions subsection would improve clarity.
  2. [Methods] No mention is made of how reward functions are designed for the diverse GRN and RL tasks, nor of any baseline comparisons against direct LLM prompting or other interface methods.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive and insightful comments. We address each major point below, clarifying our claims and indicating revisions that will strengthen the manuscript without misrepresenting the work.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the assertion that the framework 'yields fluent dialogue without altering any system parameter' and 'shows that well-trained agents of disparate origin converge on similar behavior' is presented without any quantitative metrics, success rates, example trajectories, or statistical comparisons. Because these outcomes are load-bearing for the central claim that the non-human system speaks in its own voice, the absence of supporting data or analysis undermines evaluation of the result.

    Authors: We agree that the abstract would be strengthened by including quantitative support for these claims. The main text reports success rates, behavioral convergence metrics, example trajectories, and statistical comparisons across GRN architectures and RL tasks. In revision we will add concise quantitative highlights (e.g., mean success rate and convergence measure) directly into the abstract so that the central claims can be evaluated at a glance. revision: yes

  2. Referee: [Framework description] Framework description (abstract and pipeline): the LLM is described as routing the prompt to a game whose reward semantics match and then synthesizing an environmental state 'for which the desired action is the rational response.' This construction appears to pre-encode the intended semantics in the choice of game and state, so that the observed trajectory is an elicited rather than open-ended response. A concrete ablation or control experiment isolating the system's contribution from the LLM's preparatory work is required to support the claim that meaning emerges solely from the system's inductive bias and the game.

    Authors: The LLM performs only two limited functions: (1) mapping the human prompt to an existing game whose reward function aligns with the prompt's intent, and (2) generating an initial environmental state that is consistent with the game's rules. All subsequent dynamics, state transitions, and reward accumulation are produced exclusively by the frozen nonlinear system. Meaning therefore arises from the system's own optimization of that reward, which depends on its specific inductive biases. Nevertheless, we acknowledge that an explicit control would make this separation clearer. In the revision we will add an ablation comparing performance with the trained linear interfaces against a version using random or fixed interfaces, thereby isolating the contribution of the system's native dynamics. revision: yes

Circularity Check

1 steps flagged

LLM game routing and state design pre-selects rational responses, so meaning acquisition reduces to external construction rather than emerging from frozen dynamics

specific steps
  1. self definitional [Abstract]
    "Given a human prompt, a language model routes it to the game whose semantics best match it and designs an environmental state for which the desired action is the rational response, letting the system reply through its own behavior."

    The game and state are constructed precisely so the desired action is the rational response; the system's 'reply' is then the behavior that was pre-defined as optimal. This makes acquisition of meaning equivalent to the external design choices rather than an emergent property of playing the game with the frozen dynamics.

full rationale

The paper claims meaning arises through use in the game with frozen system dynamics as the nonlinear core. However, the central pipeline explicitly uses an external LLM to select the game and synthesize a state in which the target behavior is already the rational response under the reward. This makes the observed reply an elicited trajectory by construction, reducing the 'speaks in its own voice' claim to the preparatory choices rather than an independent inductive bias of the reservoir. The derivation chain therefore contains a self-definitional step at the interface between human prompt and system output.

Axiom & Free-Parameter Ledger

1 free parameters · 2 axioms · 1 invented entities

The framework rests on philosophical assumptions about meaning, the feasibility of linear interfaces, and the ability of an LM to perform accurate routing and state design.

free parameters (1)
  • weights of linear input and output interfaces
    Trained via reinforcement learning to connect to the frozen nonlinear core
axioms (2)
  • domain assumption Meaning is located in use (Wittgenstein)
    Invoked to argue that system states acquire meaning through reward-based game participation
  • domain assumption Different architectures optimizing the same reward yield interpretable behaviors
    Used to position the game as a lingua franca across systems
invented entities (1)
  • game as lingua franca no independent evidence
    purpose: Common semantic layer allowing cross-architecture interpretation
    Introduced to unify behaviors of disparate systems under one reward

pith-pipeline@v0.9.0 · 5825 in / 1418 out tokens · 73738 ms · 2026-05-21T00:45:12.206597+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

94 extracted references · 94 canonical work pages · 4 internal anchors

  1. [1]

    Language of fungi derived from their electrical spiking activity.Royal Society Open Science, 9(4):211926, 2022

    Andrew Adamatzky. Language of fungi derived from their electrical spiking activity.Royal Society Open Science, 9(4):211926, 2022

    work page 2022

  2. [2]

    Transcription-based circadian mechanism controls the duration of molecular clock states in response to signaling inputs.Journal of Theoretical Biology, 484:110015, 2020

    Sofia Almeida, Madalena Chaves, and Franck Delaunay. Transcription-based circadian mechanism controls the duration of molecular clock states in response to signaling inputs.Journal of Theoretical Biology, 484:110015, 2020

    work page 2020

  3. [3]

    On having no head: Cognition throughout biological systems.Frontiers in Psychology, 7:902, 2016

    Franti ˇsek Baluˇska and Michael Levin. On having no head: Cognition throughout biological systems.Frontiers in Psychology, 7:902, 2016

    work page 2016

  4. [4]

    Miller, and Arthur S

    Franti ˇsek Baluˇska, William B. Miller, and Arthur S. Reber. Cellular and evolutionary perspectives on organismal cognition: from unicellular to multicellular organisms.Biological Journal of the Linnean Society, 2022

    work page 2022

  5. [5]

    The organic codes

    Marcello Barbieri. The organic codes. the basic mechanism of macroevolution.Rivista di Biologia, 91(3):481– 513, 1998

    work page 1998

  6. [6]

    Biology with information and meaning.History and Philosophy of the Life Sciences, 25(2):243–254, 2003

    Marcello Barbieri. Biology with information and meaning.History and Philosophy of the Life Sciences, 25(2):243–254, 2003

    work page 2003

  7. [7]

    Biosemiotics: a new understanding of life.Naturwissenschaften, 95(7):577–599, 2008

    Marcello Barbieri. Biosemiotics: a new understanding of life.Naturwissenschaften, 95(7):577–599, 2008

    work page 2008

  8. [8]

    What is code biology?BioSystems, 164:1–10, 2018

    Marcello Barbieri. What is code biology?BioSystems, 164:1–10, 2018

    work page 2018

  9. [9]

    Modeling feedback loops of the mammalian circadian oscillator.Biophysical Journal, 87(5):3023–3034, 2004

    Sabine Becker-Weimann, Jana Wolf, Hanspeter Herzel, and Achim Kramer. Modeling feedback loops of the mammalian circadian oscillator.Biophysical Journal, 87(5):3023–3034, 2004

    work page 2004

  10. [10]

    Bellemare, Yavar Naddaf, Joel Veness, and Michael Bowling

    Marc G. Bellemare, Yavar Naddaf, Joel Veness, and Michael Bowling. The Arcade Learning Environment: An evaluation platform for general agents.Journal of Artificial Intelligence Research, 47:253–279, 2013

    work page 2013

  11. [11]

    Estimating or Propagating Gradients Through Stochastic Neurons for Conditional Computation

    Yoshua Bengio, Nicholas L ´eonard, and Aaron Courville. Estimating or propagating gradients through stochastic neurons for conditional computation.arXiv preprint arXiv:1308.3432, 2013

    work page internal anchor Pith review Pith/arXiv arXiv 2013

  12. [12]

    The nocebo effect: history and physiology.Preventive Medicine, 26(5):612–615, 1997

    Herbert Benson. The nocebo effect: history and physiology.Preventive Medicine, 26(5):612–615, 1997

    work page 1997

  13. [13]

    Surama Biswas, Wesley Clawson, and Michael Levin. Learning in transcriptional network models: Compu- tational discovery of pathway-level memory and effective interventions.International Journal of Molecular Sciences, 24(1):285, 2022

    work page 2022

  14. [14]

    Gene regulatory networks exhibit several kinds of memory: Quantification of memory in biological and random transcriptional networks.iScience, 24(3):102131, 2021

    Surama Biswas, Santosh Manicka, Erik Hoel, and Michael Levin. Gene regulatory networks exhibit several kinds of memory: Quantification of memory in biological and random transcriptional networks.iScience, 24(3):102131, 2021

    work page 2021

  15. [15]

    Black, Georgia Christodoulou, and Steven Cole

    David S. Black, Georgia Christodoulou, and Steven Cole. Mindfulness meditation and gene expression: a hypothesis-generating framework.Current Opinion in Psychology, 28:302–306, 2019

    work page 2019

  16. [16]

    Ivana Buric, Miguel Farias, Jonathan Jong, Catherine Mee, and Inti A. Brazil. What is the molecular signature of mind-body interventions? a systematic review of gene expression changes induced by meditation and related practices.Frontiers in Immunology, 8:670, 2017

    work page 2017

  17. [17]

    Walter B. Cannon. V oodoo death.Psychosomatic Medicine, 19(3):182–190, 1957

    work page 1957

  18. [18]

    A computational model for understanding stem cell, trophectoderm and endoderm lineage determination.PLoS ONE, 3(10):e3478, 2008

    Vijay Chickarmane and Carsten Peterson. A computational model for understanding stem cell, trophectoderm and endoderm lineage determination.PLoS ONE, 3(10):e3478, 2008

    work page 2008

  19. [19]

    Nuber, Herbert M

    Vijay Chickarmane, Carl Troein, Ulrike A. Nuber, Herbert M. Sauro, and Carsten Peterson. Transcriptional dynamics of the embryonic stem cell switch.PLoS Computational Biology, 2(9):e123, 2006

    work page 2006

  20. [20]

    Ecosystem-based reservoir computing

    Alessandro Chiolerio, Zoran Konkoli, and Andrew Adamatzky. Ecosystem-based reservoir computing. hypoth- esis paper.BioSystems, 255:105525, 2025

    work page 2025

  21. [21]

    Clawson and Michael Levin

    Wesley P. Clawson and Michael Levin. Endless forms most beautiful 2.0: teleonomy and the bioengineering of chimaeric and synthetic organisms.Biological Journal of the Linnean Society, 139:457–486, 2023. 11 Language Game: Talking to Non-Human SystemsA PREPRINT

    work page 2023

  22. [22]

    Veres, and Peter Tompa

    Peter Csermely, Nevena Kunsic, Petra Mendik, Mark Kerestely, Tamas Farago, Daniel V . Veres, and Peter Tompa. Learning of signaling networks: Molecular mechanisms.Trends in Biochemical Sciences, 45(4):284–294, 2020

    work page 2020

  23. [23]

    Synthetic morphology with agential materials.Nature Reviews Bioengineering, 1:46–59, 2023

    Jamie Davies and Michael Levin. Synthetic morphology with agential materials.Nature Reviews Bioengineering, 1:46–59, 2023

    work page 2023

  24. [24]

    Gymnasium robotics, 2024

    Rodrigo de Lazcano, Andreas Kallinteris, Jun Jet Tai, Seungjae Ryan Lee, and Jordan Terry. Gymnasium robotics, 2024

    work page 2024

  25. [25]

    OCAtari: Object- centric Atari 2600 reinforcement learning environments

    Quentin Delfosse, Jannis Bl ¨uml, Bjarne Gregori, Sebastian Sztwiertnia, and Kristian Kersting. OCAtari: Object- centric Atari 2600 reinforcement learning environments. InReinforcement Learning Conference (RLC), 2024

    work page 2024

  26. [26]

    Dossett, Gregory L

    Michelle L. Dossett, Gregory L. Fricchione, and Herbert Benson. A new era for mind-body medicine.New England Journal of Medicine, 382(15):1390–1391, 2020

    work page 2020

  27. [27]

    Andrea W. M. Evers, Luana Colloca, Charlotte Blease, Marco Annoni, Lauren Y . Atlas, Fabrizio Benedetti, Ul- rike Bingel, Christian Buchel, Claudia Carvalho, Ben Colagiuri, Alia J. Crum, Paul Enck, Jens Gaab, Andrew L. Geers, Jeremy Howick, Karin B. Jensen, Irving Kirsch, Karin Meissner, Vitaly Napadow, Karin J. Peerdeman, Amir Raz, Winfried Rief, Lene Va...

    work page 2018

  28. [28]

    How do living systems create meaning?Philosophies, 5(4):36, 2020

    Chris Fields and Michael Levin. How do living systems create meaning?Philosophies, 5(4):36, 2020

    work page 2020

  29. [29]

    Thoughts and thinkers: On the complementarity between objects and processes

    Chris Fields and Michael Levin. Thoughts and thinkers: On the complementarity between objects and processes. Physics of Life Reviews, 2025

    work page 2025

  30. [30]

    Gardner, Charles R

    Timothy S. Gardner, Charles R. Cantor, and James J. Collins. Construction of a genetic toggle switch inEs- cherichia coli.Nature, 403(6767):339–342, 2000

    work page 2000

  31. [31]

    Gardner, Milos Dolnik, and James J

    Timothy S. Gardner, Milos Dolnik, and James J. Collins. A theory for controlling cell cycle dynamics using a reversibly binding inhibitor.Proceedings of the National Academy of Sciences, 95(24):14190–14195, 1998

    work page 1998

  32. [32]

    A skeleton model for the network of cyclin-dependent kinases driving the mammalian cell cycle.Interface Focus, 1(1):24–35, 2011

    Claude G ´erard and Albert Goldbeter. A skeleton model for the network of cyclin-dependent kinases driving the mammalian cell cycle.Interface Focus, 1(1):24–35, 2011

    work page 2011

  33. [33]

    Oscillations and variability in the p53 system.Molecular Systems Biology, 2:2006.0033, 2006

    Naama Geva-Zatorsky, Nitzan Rosenfeld, Shalev Itzkovitz, Ron Milo, Alex Sigal, Erez Dekel, Talia Yarnitzky, Yuvalal Liron, Paz Polak, Galit Lahav, and Uri Alon. Oscillations and variability in the p53 system.Molecular Systems Biology, 2:2006.0033, 2006

    work page arXiv 2006

  34. [34]

    Dolphingemma: How google ai is helping decode dolphin communication

    Denise Herzing and Thad Starner. Dolphingemma: How google ai is helping decode dolphin communication. https://blog.google/technology/ai/dolphingemma/, 2025. Google. Retrieved June 23, 2025

    work page 2025

  35. [35]

    The Platonic Representation Hypothesis

    Minyoung Huh, Brian Cheung, Tongzhou Wang, and Phillip Isola. The platonic representation hypothesis.arXiv preprint arXiv:2405.07987, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  36. [36]

    Harnessing nonlinearity: Predicting chaotic systems and saving energy in wireless communication.science, 304(5667):78–80, 2004

    Herbert Jaeger and Harald Haas. Harnessing nonlinearity: Predicting chaotic systems and saving energy in wireless communication.science, 304(5667):78–80, 2004

    work page 2004

  37. [37]

    Kamm, Rashid Bashir, Nitin Arora, Roy D

    Roger D. Kamm, Rashid Bashir, Nitin Arora, Roy D. Dar, Martha U. Gillette, Linda G. Griffith, Melissa L. Kemp, Kathy Kinlaw, Michael Levin, Adam C. Martin, Todd C. McDevitt, Robert M. Nerem, Michael J. Pow- ers, Taher A. Saif, James Sharpe, Shuichi Takayama, Shoji Takeuchi, Ron Weiss, Keith M. Ye, Hannah G. Yevick, and Muhammad H. Zaman. Perspective: The ...

    work page 2018

  38. [38]

    Kauffman.Reinventing the Sacred: A New View of Science, Reason and Religion

    Stuart A. Kauffman.Reinventing the Sacred: A New View of Science, Reason and Religion. Basic Books, New York, 2008

    work page 2008

  39. [39]

    K ¨ubra Kaygisiz and Rein V . Ulijn. Can molecular systems learn?ChemSystemsChem, 7:e202400075, 2024

    work page 2024

  40. [40]

    Kovacs, Klara Schulc, Daniel V

    David Keresztes, Mark Kerestely, Levente Szarka, Borbala M. Kovacs, Klara Schulc, Daniel V . Veres, and Pe- ter Csermely. Cancer drug resistance as learning of signaling networks.Biomedicine & Pharmacotherapy, 183:117880, 2025

    work page 2025

  41. [41]

    Kholodenko

    Boris N. Kholodenko. Negative feedback and ultrasensitivity can bring about oscillations in the mitogen-activated protein kinase cascades.European Journal of Biochemistry, 267(6):1583–1588, 2000

    work page 2000

  42. [42]

    Future medicine: from molecular pathways to the collective intelligence of the body.Trends in Molecular Medicine, 2023

    Evan Lagasse and Michael Levin. Future medicine: from molecular pathways to the collective intelligence of the body.Trends in Molecular Medicine, 2023

    work page 2023

  43. [43]

    Multi-agent cooperation and the emergence of (natural) language

    Angeliki Lazaridou, Alexander Peysakhovich, and Marco Baroni. Multi-agent cooperation and the emergence of (natural) language. InInternational Conference on Learning Representations, 2017. 12 Language Game: Talking to Non-Human SystemsA PREPRINT

    work page 2017

  44. [44]

    Chaos and birhythmicity in a model for circadian oscillations of the PER and TIM proteins inDrosophila.Journal of Theoretical Biology, 198(3):445–459, 1999

    Jean-Christophe Leloup and Albert Goldbeter. Chaos and birhythmicity in a model for circadian oscillations of the PER and TIM proteins inDrosophila.Journal of Theoretical Biology, 198(3):445–459, 1999

    work page 1999

  45. [45]

    Bioelectric signaling: Reprogrammable circuits underlying embryogenesis, regeneration, and cancer.Cell, 184(8):1971–1989, 2021

    Michael Levin. Bioelectric signaling: Reprogrammable circuits underlying embryogenesis, regeneration, and cancer.Cell, 184(8):1971–1989, 2021

    work page 1971

  46. [46]

    Technological approach to mind everywhere: An experimentally-grounded framework for un- derstanding diverse bodies and minds.Frontiers in Systems Neuroscience, 16:768201, 2022

    Michael Levin. Technological approach to mind everywhere: An experimentally-grounded framework for un- derstanding diverse bodies and minds.Frontiers in Systems Neuroscience, 16:768201, 2022

    work page 2022

  47. [47]

    Bioelectric networks: the cognitive glue enabling evolutionary scaling from physiology to mind

    Michael Levin. Bioelectric networks: the cognitive glue enabling evolutionary scaling from physiology to mind. Animal Cognition, 2023

    work page 2023

  48. [48]

    The multiscale wisdom of the body: Collective intelligence as a tractable interface for next- generation biomedicine.BioEssays, page e202400196, 2024

    Michael Levin. The multiscale wisdom of the body: Collective intelligence as a tractable interface for next- generation biomedicine.BioEssays, page e202400196, 2024

    work page 2024

  49. [49]

    Self-improvising memory: A perspective on memories as agential, dynamically reinterpreting cognitive glue.Entropy, 26(6):481, 2024

    Michael Levin. Self-improvising memory: A perspective on memories as agential, dynamically reinterpreting cognitive glue.Entropy, 26(6):481, 2024

    work page 2024

  50. [50]

    Artificial intelligences: A bridge toward diverse intelligence and humanity’s future.Advanced Intelligent Systems, page 2401034, 2025

    Michael Levin. Artificial intelligences: A bridge toward diverse intelligence and humanity’s future.Advanced Intelligent Systems, page 2401034, 2025

    work page 2025

  51. [51]

    Cell2sentence: Teaching large language models the language of biology

    Daniel Levine, Syed A Rizvi, Sacha L ´evy, Nazreen Pallikkavaliyaveetil, David Zhang, Xingyu Chen, Sina Ghadermarzi, Ruiming Wu, Zihe Zheng, Ivan Vrkic, Anna Zhong, Daphne Raskin, Insu Han, Antonio Hen- rique de Oliveira Fonseca, Josue Ortega Caro, Amin Karbasi, Rahul Madhav Dhodapkar, and David van Dijk. Cell2sentence: Teaching large language models the ...

    work page 2024

  52. [52]

    Harvard University Press, 1969

    David Lewis.Convention: A Philosophical Study. Harvard University Press, 1969

    work page 1969

  53. [53]

    Liebal, Praveen K

    Ulf W. Liebal, Praveen K. Sappa, Thomas Millat, Leif Steil, Georg Homuth, Uwe V ¨olker, and Olaf Wolken- hauer. Proteolysis ofβ-galactosidase following SigmaB activation inBacillus subtilis.Molecular BioSystems, 8(6):1806–1814, 2012

    work page 2012

  54. [54]

    ODEbase: a repository of ODE systems for systems biology.Bioinformatics Advances, 2(1):vbac027, 2022

    Christoph L ¨uders, Thomas Sturm, and Ovidiu Radulescu. ODEbase: a repository of ODE systems for systems biology.Bioinformatics Advances, 2(1):vbac027, 2022

    work page 2022

  55. [55]

    The biogenic approach to cognition.Cognitive Processing, 7(1):11–29, 2006

    Pamela Lyon. The biogenic approach to cognition.Cognitive Processing, 7(1):11–29, 2006

    work page 2006

  56. [56]

    The cognitive cell: bacterial behavior reconsidered.Frontiers in Microbiology, 6:264, 2015

    Pamela Lyon. The cognitive cell: bacterial behavior reconsidered.Frontiers in Microbiology, 6:264, 2015

    work page 2015

  57. [57]

    Real-time computing without stable states: A new framework for neural computation based on perturbations.Neural computation, 14(11):2531–2560, 2002

    Wolfgang Maass, Thomas Natschl ¨ager, and Henry Markram. Real-time computing without stable states: A new framework for neural computation based on perturbations.Neural computation, 14(11):2531–2560, 2002

    work page 2002

  58. [58]

    Signaling switches and bistability arising from multisite phosphorylation in protein kinase cascades.The Journal of cell biology, 164(3):353–359, 2004

    Nick I Markevich, Jan B Hoek, and Boris N Kholodenko. Signaling switches and bistability arising from multisite phosphorylation in protein kinase cascades.The Journal of cell biology, 164(3):353–359, 2004

    work page 2004

  59. [59]

    Cellular signaling pathways as plastic, proto-cognitive systems: Implications for biomedicine.Patterns, 4(5):100737, 2023

    Juanita Mathews, Alan Jaelyn Chang, Liam Devlin, and Michael Levin. Cellular signaling pathways as plastic, proto-cognitive systems: Implications for biomedicine.Patterns, 4(5):100737, 2023

    work page 2023

  60. [60]

    Emergence of grounded compositional language in multi-agent populations

    Igor Mordatch and Pieter Abbeel. Emergence of grounded compositional language in multi-agent populations. InProceedings of the AAAI conference on artificial intelligence, volume 32, 2018

    work page 2018

  61. [61]

    Multi-scale chimerism: An experimental window on the algorithms of anatomical control.Cells & Development, 169:203764, 2021

    Vasilios Nanos and Michael Levin. Multi-scale chimerism: An experimental window on the algorithms of anatomical control.Cells & Development, 169:203764, 2021

    work page 2021

  62. [62]

    Hadad, and David Omer

    Guy Oren, Aner Shapira, Reuven Lifshitz, Ehud Vinepinsky, Roni Cohen, Tomer Fried, Guy P. Hadad, and David Omer. V ocal labeling of others by nonhuman primates.Science, 385(6712):996–1003, 2024

    work page 2024

  63. [63]

    Osgood and Thomas A

    Charles E. Osgood and Thomas A. Sebeok. Psycholinguistics: a survey of theory and research problems.Journal of Abnormal Psychology, 49(4):1–203, 1954

    work page 1954

  64. [64]

    Pardo, Kurt Fristrup, David S

    Michael A. Pardo, Kurt Fristrup, David S. Lolchuragi, Joyce H. Poole, Petter Granli, Cynthia Moss, Iain Douglas- Hamilton, and George Wittemyer. African elephants address one another with individually specific name-like calls.Nature Ecology & Evolution, 8:1353–1364, 2024

    work page 2024

  65. [65]

    Howard H. Pattee. Cell psychology: An evolutionary approach to the symbol-matter problem.Cognition and Brain Theory, 5(4):325–341, 1982

    work page 1982

  66. [66]

    Howard H. Pattee. The measurement problem in artificial world models.BioSystems, 23(2–3):281–290, 1989

    work page 1989

  67. [67]

    Howard H. Pattee. The physics of symbols: bridging the epistemic cut.BioSystems, 60(1–3):5–21, 2001

    work page 2001

  68. [68]

    Charles S. Peirce. The charles s. peirce papers. Annotated Catalogue of the Papers of Charles S. Peirce by Richard S. Robin, 1967. 13 Language Game: Talking to Non-Human SystemsA PREPRINT

    work page 1967

  69. [69]

    A review of the model/rival (m/r) technique for training interspecies communication and its use in behavioral research.Animals, 11(9):2479, 2021

    Irene M Pepperberg. A review of the model/rival (m/r) technique for training interspecies communication and its use in behavioral research.Animals, 11(9):2479, 2021

    work page 2021

  70. [70]

    Words and drugs: Same mechanisms of action?Journal of Contemporary Psychotherapy, 46(3):159–166, 2016

    Alessandro Piedimonte and Fabrizio Benedetti. Words and drugs: Same mechanisms of action?Journal of Contemporary Psychotherapy, 46(3):159–166, 2016

    work page 2016

  71. [71]

    Harnessing the analog computing power of regulatory networks with the regulatory network machine.iScience, 28(6):112536, 2025

    Alexis Pietak and Michael Levin. Harnessing the analog computing power of regulatory networks with the regulatory network machine.iScience, 28(6):112536, 2025

    work page 2025

  72. [72]

    Reber and Franti ˇsek Balu ˇska

    Arthur S. Reber and Franti ˇsek Balu ˇska. Cognition in some surprising places.Biochemical and Biophysical Research Communications, 564:150–157, 2021

    work page 2021

  73. [73]

    NatureLM-audio: An audio-language foundation model for bioacoustics

    David Robinson, Marius Miron, Masato Hagiwara, Benno Weck, Sara Keen, Milad Alizadeh, Gagan Narula, Matthieu Geist, and Olivier Pietquin. NatureLM-audio: An audio-language foundation model for bioacoustics. InInternational Conference on Learning Representations, 2025

    work page 2025

  74. [74]

    Pergamon Press, Oxford, 1985

    Robert Rosen.Anticipatory Systems: Philosophical, Mathematical, and Methodological Foundations. Pergamon Press, Oxford, 1985

    work page 1985

  75. [75]

    The multiple realizability of sentience in living systems and beyond

    Nicolas Rouleau and Michael Levin. The multiple realizability of sentience in living systems and beyond. eNeuro, 10(11), 2023

    work page 2023

  76. [76]

    Herbert benson: mind-body maverick pushes the envelope.Science, 276(5311):357–359, 1997

    Wade Roush. Herbert benson: mind-body maverick pushes the envelope.Science, 276(5311):357–359, 1997

    work page 1997

  77. [77]

    Vernes, Katherine Zacarian, and Dami ´an E

    Christian Rutz, Michael Bronstein, Aza Raskin, Sonja C. Vernes, Katherine Zacarian, and Dami ´an E. Blasi. Using machine learning to decode animal communication.Science, 381(6654):152–155, 2023

    work page 2023

  78. [78]

    Proximal Policy Optimization Algorithms

    John Schulman, Filip Wolski, Prafulla Dhariwal, Alec Radford, and Oleg Klimov. Proximal policy optimization algorithms.arXiv preprint arXiv:1707.06347, 2017

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  79. [79]

    Thomas A. Sebeok. Animal communication.Science, 147(3661):1006–1014, 1965

    work page 1965

  80. [80]

    C. Shang. Emerging paradigms in mind-body medicine.Journal of Alternative and Complementary Medicine, 7(1):83–91, 2001

    work page 2001

Showing first 80 references.