# "On Biological and Artificial Consciousness" by Borjan Milinkovic and Jaan Aru Source: https://www.youtube.com/watch?v=KKrRLgPp6XI Recap page: https://rapidrecap.app/video/KKrRLgPp6XI Generated: 2026-03-05T16:07:49.062+00:00 --- ## Quick Overview Borjan Milinkovic argues for biological computationalism, asserting that consciousness is not solely dependent on abstract computational organization (like in digital systems) but requires specific biological, multi-scale, and continuous dynamics, such as ephaptic coupling and scale-integration, which digital systems currently lack. **Key Points:** - The presentation advocates for biological computationalism, arguing that consciousness arises from biological processes that are fundamentally different from standard digital computation. - Digital computation is characterized by scale-separability, algorithmic closure, and dependence only on discrete states (e.g., binary), unlike biological systems. - Biological computation involves scale-integration across multiple scales (microcolumn, field effects, macro-scale brain dynamics), leading to a heterarchy rather than a strict hierarchy. - Metabolic constraints, such as energy limits and ATP efficiency tuning of ion channels, structurally and dynamically shape neural interactions. - Dendrites play a crucial role by blurring digital boundaries through local non-linear integration and back-propagating spikes, meaning a single biological neuron is not simply a perceptron neuron. - The core question shifts from 'Can synthetic systems be conscious?' to 'Are we building the right machines?' that can realize these hybrid, scale-integrated, and dynamically constrained computations. - The presentation includes references to work on ion-channel kinetics, graded potentials, and BOLD signals as active areas of research supporting these biological constraints. ![Screenshot at 00:23: The slide explicitly asks, "Can synthetic systems be conscious, and what would that require?" followed by the idea of building a 'Phenomenal Engine,' setting the stage for the argument that current digital computation paradigms are insufficient.](https://ss.rapidrecap.app/screens/KKrRLgPp6XI/00-00-23.jpg) **Context:** This presentation, titled "On Biological and Artificial Consciousness: A case for biological computationalism," is delivered by Borjan Milinkovic (Boki) of the Paris-Saclay Institute of Neurosciences (NeuroPSI), CNRS, as part of the Michael Levin Lab Talk series. The talk challenges the traditional computational functionalism view of consciousness, which posits that consciousness is substrate-independent and purely algorithmic, by arguing that biological consciousness relies on specific physical and multi-scale dynamics inherent to living systems. ## Detailed Analysis The presentation argues against pure computational functionalism in favor of biological computationalism. Milinkovic establishes the debate on consciousness as existing on a spectrum between two camps: Computational Functionalism (which holds that consciousness is substrate-independent and relies only on information processing) and Biological Naturalism (which insists on the necessity of specific biological substrate/processes). The speaker places his work in the middle, arguing that consciousness depends on specific biological dynamics that are neither purely algorithmic nor purely physical in the digital sense. He introduces three pillars supporting biological computationalism. Pillar 1 focuses on metabolic constraints, stating that energy limits shape neural interactions structurally and dynamically, citing research on ATP efficiency tuning of ion channels. Pillar 2 addresses Multiscale organization and scale-integration, emphasizing a heterarchy of distributed constraints rather than a hierarchy of control, meaning no single scale (molecules, neurons, regions) is privileged. This scale-integration is crucial for conscious processing. Pillar 3 focuses on hybrid continuous-discrete computation, noting that spikes are not the whole story, as dendrites perform local non-linear integration and back-propagating spikes, blurring the digital boundaries often assumed. The conclusion is that while digital systems approximate continuity, biological systems instantiate it, suggesting that building artificial consciousness requires machines that respect these biologically informed, multi-scale, continuous, and dynamically constrained dynamics. ### Introduction and Core Argument - Milinkovic introduces the core tension between biological and artificial consciousness, favoring a view where biological substrate matters. ### Pillar 1 - Metabolic constraints shape dynamics and computations - Energy limits are not peripheral; they shape neural interactions structurally and dynamically - Ion-channel kinetics are tuned for ATP efficiency (Hasenstaud et al. 2010). ### Pillar 1 (cont.) - Coarse-graining induces better information transmission under constraints - Non-spiking graded potentials are energetically cheaper than spikes (Magistretti and Allaman, 2015). ### Pillar 2 - Multiscale organisation and scale-integration: Heterarchy not hierarchy - No single scale is in charge structurally (not molecules, neurons, regions) or functionally. ### Pillar 2 (cont.) - Scale-integration emerges as an optimization strategy, leading to a heterarchy of distributed constraint, not hierarchy of control. ### Pillar 3 - Hybrid continuous-discrete computation - Spikes are not the whole story; dendrites blur digital boundaries via local non-linear integration and timing-sensitive computation. ### Conclusion - Consciousness relies on scale-integration and hybrid dynamics, suggesting current digital architectures (von Neumann) are fundamentally incomplete for realizing consciousness. ![Screenshot at 00:09: Speaker Borjan Milinkovic introduces the topic by contrasting biological and artificial consciousness.](https://ss.rapidrecap.app/screens/KKrRLgPp6XI/00-00-09.jpg) ![Screenshot at 03:24: Slide posing the central question: "Can synthetic systems be conscious, and what would that require?" leading to the idea of a "Phenomenal Engine."](https://ss.rapidrecap.app/screens/KKrRLgPp6XI/00-03-24.jpg) ![Screenshot at 06:14: Slide contrasting Computational Functionalism \(left, image of a computer-like structure\) and Biological Naturalism \(right, image of Frankenstein's monster\).](https://ss.rapidrecap.app/screens/KKrRLgPp6XI/00-06-14.jpg) ![Screenshot at 08:38: Slide summarizing the premises of Computational Functionalism: consciousness supervenes on computational organization, and substrate-independence suggests neurons, carbon, silicon, or any medium are equivalent.](https://ss.rapidrecap.app/screens/KKrRLgPp6XI/00-08-38.jpg) ![Screenshot at 22:29: Slide contrasting digital computation \(decomposable, modular, separable\) with biological computation \(none of these\), emphasizing metabolic constraints on biological systems.](https://ss.rapidrecap.app/screens/KKrRLgPp6XI/00-22-29.jpg)