The question of "what is life?" has challenged scientists and philosophers for centuries, producing an array of definitions that reflect both the mystery of its emergence and the diversity of disciplinary perspectives brought to bear on the question. Despite significant progress in our understanding of biological systems, psychology, computation, and information theory, no single definition for life has yet achieved universal acceptance. This challenge becomes increasingly urgent as advances in synthetic biology, artificial intelligence, and astrobiology challenge our traditional conceptions of what it means to be alive. We undertook a methodological approach that leverages large language models (LLMs) to analyze a set of definitions of life provided by a curated set of cross-disciplinary experts. We used a novel pairwise correlation analysis to map the definitions into distinct feature vectors, followed by agglomerative clustering, intra-cluster semantic analysis, and t-SNE (t-distributed Stochastic Neighbor Embedding) projection to reveal underlying conceptual archetypes. This methodology revealed a continuous landscape of the themes relating to the definition of life, suggesting that what has historically been approached as a binary taxonomic problem should be instead conceived as differentiated perspectives within a unified conceptual latent space. We offer a new methodological bridge between reductionist and holistic approaches to fundamental questions in science and philosophy, demonstrating how computational semantic analysis can reveal conceptual patterns across disciplinary boundaries, and opening similar pathways for addressing other contested definitional territories across the sciences.

The problem of aligning humans and artificial intelligences can be understood in terms of minimizing externalities between them. However, economics cannot define externality because it contradicts the rationality assumption. This paper applies the homeostatic principles, from anatomical homeostasis to its disorder – cancer, to define externality. Drawing upon the perspective of cancer as a problem of scaling cellular collectives, this paper shows how to redefine both externality and rationality in terms of cognitive light cones (which demarcate the scale of goals any agent can pursue). We propose that cognitive light cones are constructed out of interoceptive signals for the purpose of anatomical homeostasis. We show that externalities can be understood in terms of anatomical homeostasis and derive some important implications for AI alignment, including the possibility of using market mechanisms enable the mutual co-construction of alignment between artificial intelligences and humans.

Background: Colorectal cancer (CRC) remains a leading cause of cancer mortality and a promising target for emerging “ionoceutical” therapies that modulate cancer channelopathies. Ion channels are key regulators of tumor bioelectric signaling, yet their biomarker and therapeutic potential in CRC are only beginning to emerge from preclinical and transcriptomic studies. Methods: Public RNA‑seq datasets from primary CRC samples and COLO 205 cells were compared with normal colon tissue to identify shared upregulation of ion channel genes. Guided by these findings, 19 ion‑modulating compounds targeting dysregulated channels were screened in COLO 205 reporter lines. Top hits were evaluated for effects on proliferation and cytotoxicity in COLO 205 (p53 wild‑type), HT‑29 (p53 mutant), and Caco‑2 (p53 null) cells, with cell‑cycle distribution analyzed in COLO 205. Activity, selectivity, and in vivo efficacy were assessed using 2D and 3D COLO 205 spheroid assays, heterotypic COLO 205–endothelial–fibroblast intravasation spheroids, non‑cancerous toxicity assays in normal colon epithelial, endothelial, and colon fibroblast cells, and in vivo COLO 205 xenografts. Mechanism was probed using immunoblotting and RNA‑seq. Results: Nine compounds inhibited CRC cell proliferation primarily through cytostatic mechanisms (G0/G1 accumulation). Most agents were less active in 3D spheroids than in 2D; however, the potassium‑channel, Na+/K+-ATPase pump, and NMDA receptor antagonist clofilium had greatly enhanced efficacy, impaired migration and invasion, and showed the highest selectivity (IC50 = 3.317 µM in COLO 205; therapeutic index = 47 vs normal colon epithelial cells). Clofilium induced acute depolarization followed by sustained hyperpolarization, caused broad transcriptional reprogramming, ER stress, cell‑cycle arrest, and innate immunity pathway activation, with immunoblotting confirming inhibition of AKT, JNK, p38, and NF‑κB and induction of NDRG1, senescence, autophagy, and caspase‑3 cleavage. In heterotypic spheroids, clofilium reduced intravasation and exerted anti‑angiogenic effects, and in COLO 205 xenografts daily treatment reduced tumor volume by 37% with minimal systemic toxicity. Conclusions: Clofilium demonstrates potent, selective anti‑CRC activity and supports ion channel targeting as a viable “ionoceutical” strategy for CRC.

Intelligence appears under different names in different fields: as data compression in statistics and machine learning, as universal computation in dynamical systems, and as adaptive behavior in agents. Each field carries its own objective, and the two most influential drives often fail in mirror image: novelty search, which seeks surprise, is transfixed by a noisy television screen, while the free-energy principle, which avoids surprise, is most content in a dark room. Both failures have a single cause: each objective treats as one quantity the surprise a learner can convert into knowledge and the surprise it never can. Here we show that the learnable part of that information, which we call learnable novelty, yields the seemingly disparate projections of intelligence, and we give a closed-form estimator of it built on a cheap and differentiable reservoir computer. Used as a measure, with no supervision of any kind, the estimator recovers decades of complexity classification, ranking the Turing-complete rule 110 highest among the elementary cellular automata. Used as an objective, its gradient carries a neural cellular automaton from simple dynamics into a regime of solitons, the traveling, colliding structures by which rule 110 computes, as well as organizes the representation of an image encoder around the ten digit classes of MNIST, fully unsupervised: no label ever enters training. Handed to a reinforcement-learning agent as an intrinsic reward, it supplies the exploration that task rewards lack, improving on the task baseline in nine of ten environments and collapsing in none. Complexity generation, abstraction, and exploration, ordinarily pursued with unrelated objectives in separate fields, thus emerge from ascent on one differentiable quantity, and the projections of intelligence gain a common quantitative footing.

Abstract This special issue examines how natural and artificial intelligences (AIs) model the world, and what this modelling reveals about cognition and relationships between life and mind. Rather than adopting a single definition, the collection considers how world models function and emerge in biological and artificial systems, exploring a diverse range of world modelling including causal, self-referential, individual goal-directed, collective and narrative forms. A recurring theme is the extent to which current AI systems trained on vast quantities of data learn the context-sensitive, temporally embedded, value-laden dimensions of world modelling that characterize diverse biological intelligences, or whether their impressive capabilities arise primarily from statistical surface regularities. The contributions also raise broader issues concerning embodiment, complexity, learning architectures and the social and scientific contexts in which world models operate. With this collection, we hope to clarify the conceptual landscape, identify key points of similarity and divergence between natural and artificial minds, and outline questions that may guide future research on the forms of world modelling that support grounded understanding, robust agency and potentially human-like general intelligence. This article is part of the theme issue ‘World models in natural and artificial intelligence’.

Most of the Neural Cellular Automata (NCAs) defined in the literature have a common theme: they are based on regular grids with a Moore neighborhood (one-hop neighbour). They do not take into account long-range connections and more complex topologies as we can find in the brain. In this paper, we introduce BraiNCA, a brain-inspired NCA with an attention layer, long-range connections and complex topology. BraiNCAs shows better results in terms of robustness and speed of learning on the two tasks compared to Vanilla NCAs establishing that incorporating attention-based message selection together with explicit long-range edges can yield more sample-efficient and damage-tolerant self-organization than purely local, grid-based update rules. These results support the hypothesis that, for tasks requiring distributed coordination over extended spatial and temporal scales, the choice of interaction topology and the ability to dynamically route information will impact the robustness and speed of learning of an NCA. More broadly, BraiNCA provides brain-inspired NCA formulation that preserves the decentralized local update principle while better reflecting non-local connectivity patterns, making it a promising substrate for studying collective computation under biologically-realistic network structure and evolving cognitive substrates.

Achieving advanced machine intelligence remains a central challenge in AI research, often approached through scaling neural architectures and generative models. However, biological systems offer a broader repertoire of strategies for adaptive, goal-directed behavior - strategies that emerged long before nervous systems evolved. This paper advocates a genuinely life-inspired approach to machine intelligence, drawing on principles from biology that enable robustness, autonomy, and open-ended problem-solving across scales. We frame intelligence as flexible problem-solving, following William James, and develop the concept of "cognitive light cones" to characterize the continuum of intelligence in living systems and machines. We argue that biological evolution has discovered a scalable recipe for intelligence - and the progressive expansion of organisms' "cognitive light cone", predictive and control capacities. To explain how this is possible, we distill five design principles - multiscale autonomy, growth through self-assemblage of active components, continuous reconstruction of capabilities, exploitation of physical and embodied constraints, and pervasive signaling enabling self-organization and top-down control from goals - that underpin life's ability to navigate creatively diverse problem spaces. We discuss how these principles contrast with current AI paradigms and outline pathways for integrating them into future autonomous, embodied, and resilient artificial systems.

A great deal is known about the formation and architecture of biological neural networks in animal models, which have arrived at their current structure-function relationship through evolution by natural selection. Little is known about the development of such structure-function relationships in a scenario where neurons are allowed to grow within evolutionarily-novel, motile bodies. Previous work showed that ectodermal tissue excised from Xenopus embryos, develops into a three-dimensional mucociliary epidermal organoid ex vivo and exhibits movements distinct from age-matched tadpoles. These 'biobots' are autonomous, self-powered, and able to move through aqueous environments. Here, we report a new type of biobot, the neurobot, composed of mucociliary epidermis and neural tissue. We show that neural precursor cells implanted in explanted Xenopus ectodermal tissue develop into mature neurons, extending processes both toward the surface and among each other. These self-organized neurobots exhibit unique morphology, more complex movements, and different responses to neuroactive drugs compared to non-neuronal counterparts. Calcium imaging confirms neuronal activity in neurobots. Transcriptomics reveals increased transcript variability, expression of genes related to nervous system development, a shift toward ancient genes, and up-regulation of neuronal genes linked to visual perception.

Abstract Exposure to the potassium channel blocker barium chloride (BaCl₂) causes head degeneration in Dugesia japonica flatworms, followed by regeneration of BaCl 2 -insensitive heads, offering a unique model for studying transcriptional resilience to novel stress. We performed RNA sequencing on individual planaria to investigate different transcriptional solutions to the BaCl 2 challenge, and how regeneration history and social environment shape transcriptomic responses to BaCl₂. We identified a robust transcriptional strategy and a potential sub-strategy for enabling BaCl 2 -insensitive head formation. Moreover, we observed pronounced transcriptional differences between untreated worms regenerating from tail fission fragments (tail-regenerated), and untreated full-sized worms that did not fission during the experiment (intact controls), highlighting the lasting impact of regeneration history. Relative to controls, tail-regenerated worms upregulated neurodevelopmental and morphogenetic programs, while downregulating mitochondrial transport and stress-response pathways. Relative to intact controls, BaCl₂-exposed regenerates upregulated ion transport, metabolic, cell cycle, and inflammatory pathways, while downregulating neuronal signaling, ion homeostasis, morphogenesis, and tissue repair programs. Comparison of BaCl₂-exposed isolated and BaCl₂-exposed group-housed worms revealed minimal transcriptional divergence between social conditions. These findings underscore the complex interplay between regeneration, chemical stress, and social context in shaping gene expression.

Abstract The ability of self-organizing systems to display emergent, adaptive capabilities is a fundamental feature of biological life. Understanding the mechanisms by which cells co-ordinate at the micro-scale to produce macro-scale structures and behaviors is a fundamental problem in developmental biology. Moreover, it is an important goal of biomedicine to identify triggers that re-wire the physiological patterns of information flow and control signals. Here, we use a recently developed, synthetic biology platform known as basal Xenobots to explore how patterns of information flow and multi-cellular integration are regulated by chemical signaling pathways. Basal Xenobots are modified, organoid-like systems constructed from embryos, and have previously been shown to display complex patterns of information flows. In this study, we recorded calcium signals before and after exposing basal Xenobots to extra-cellular adenosine triphosphate (eATP), and used statistics from multivariate information theory to explore the differences in global patterns of information flow across the cells in each state. We found that eATP resulted in a dramatic reconfiguration of global information processing dynamics, characterized by a global decrease in multi-cellular co-ordination, reduced information transfer and integration, and a decrease in the global entropy rate of the basal Xenobots. These results provide evidence that purinergic signaling may play a key role in the regulation of multi-cellular self-organization, with implications for a variety of clinical disorders thought to involve aberrant purinergic signaling. These results also suggest the possibility that bioengineers may be able to “tune” the degree of self-organizing capacity in living systems via pharmacological intervention.

Niche construction typically refers to a set of ideas around bi-directional feedback between a species and its environment, and its impacts on the course of evolution. More fundamentally, it emphasizes the active aspect of life forms that alter their environment and establish a feedback loop in which that environment inevitably changes their behavior, structure, and future evolution. Here, we argue that this powerful dynamic is general, and extends far beyond its typical application in ecology and evolutionary biology. We start from a predictive processing view of the brain and explain how niche construction extends the scope of classical predictive and control loops beyond the nervous system and organism-level. We then use examples from cognitive science, psychopathology, cell and developmental biology, cancer biology, robotics, and AI, illustrating how agents use both living and non-living aspects of their microenvironment as a scratchpad, allowing a form of active long-term memory that supports cohesion of agency over time. Besides its memory function, niche construction enables "offloading" to the environment (externalizing) various cognitive operations, including planning, problem solving, and social coordination. We also discuss niche construction an example of the plasticity of the machine/data mapping, enabling analysis of systems from the perspective of the patterns within excitable media (agential data). Finally, we show data illustrating a new kind of niche construction: active stigmergy, in which cells leave dynamic, bioelectrically-active components in the microenvironment. By recognizing niche construction dynamics at novel spatiotemporal scales, important invariants across disciplines and substrates can be recognized and used to drive advances in biomedicine, engineering, and AI.

A hallmark of life on Earth is the ability of agents to exert causal power and be drivers of subsequent events. This is key to cognition at all scales. Causal emergence, measuring the degree to which an agent exerts unique predictive power on its future, is one consequence of causal power. Indeed, recent discoveries have shown that biological agents, even minimal ones, increase their causal emergence after learning new memories. However, there is a major knowledge gap regarding how causally emergent artificial agents are. We focused on Reinforcement Learning (RL) of neural-network agents across an array of environmental conditions, encompassing different algorithms, agent architectures, and six environments arranged on a complexity spectrum. For consistency, we computed the causal emergence of their latent-space representations over their lifetimes. We used the recently proposed ΦID to estimate causal emergence and tested how it related to learning performance. Our results suggested a Causally Emergent Alignment Hypothesis: successful agents exhibited causal emergence that was consistently predictive of final reward early in training and whose representational dynamics aligned with reward improvement in most tasks. This idea suggests that causal emergence may be a previously undisclosed axis of reorganization of neural representations in RL agents, with the potential to establish causal relationships and interventions that will lead to better RL agents. Our work also highlights the alignment between causal emergence and learning as another way biological and artificial creatures compare.

Biological systems maintain structure and function through hierarchical control, redundancy and surveillance across scales. These same features create leverage points for agents that establish a persistent presence in host systems, including foreign agents such as pathogens, parasites, trophoblasts and gall-inducing insects, and self-defecting agents such as cancer. This review synthesises evidence across immunology, developmental biology, neuroscience, oncology and ecology to treat manipulation as a staged control problem in which local access is converted into distributed effects through shared signalling and feedback loops across scales. We outline a six-step framework in which manipulators establish access, interfere with boundary and identity checks, exploit transient plasticity to reset homeostatic constraints, redirect immune, endocrine and neural communication, remodel host structure through developmental programmes, and stabilise altered states through niche construction and epigenetic remodelling. The framework separates mechanisms required for initiation from those required for maintenance and long-term persistence. We illustrate this framework through four case studies: trophoblast invasion in pregnancy, tumour progression, parasitoid manipulation by Cotesia congregata, and plant galls. The synthesis highlights recurrent vulnerabilities in boundaries, shared signalling and plasticity windows, and motivates stage-specific experiments that test whether candidate mechanisms drive entry, sustain the manipulated state, or increase resistance to reversal. The same logic can help organise thinking about complex pathology in which altered states are maintained by distributed feedback.

The emerging field of diverse intelligence seeks an integrated view of problem-solving in agents of very different provenance, composition, and substrates. From subcellular chemical networks to swarms of organisms, and across evolved, engineered, and chimeric systems, it is hypothesized that scale-invariant principles of decision-making can be discovered. We propose that cognition in both natural and synthetic systems can be characterized and understood by the interplay between two equally important invariants: (i) the remapping of embedding spaces, and (ii) the navigation within these spaces. Biological collectives, from single cells to entire organisms (and beyond), remap transcriptional, morphological, physiological, or 3D spaces to maintain homeostasis and regenerate structure, while navigating these spaces through distributed error correction. Modern Artificial Intelligence (AI) systems, including transformers, diffusion models, and neural cellular automata, enact analogous processes by remapping data into latent embeddings and refining them iteratively through contextualization. We argue that this dual principle - remapping and navigation of embedding spaces via iterative error minimization - constitutes a substrate-independent invariant of cognition. Recognizing this shared mechanism not only illuminates deep parallels between living systems and artificial models, but also provides a unifying framework for engineering adaptive intelligence across scales.

The goal of detecting, measuring, and engineering machine consciousness depends on making explicit key metaphysical assumptions about the nature of consciousness that may be lying dormant in the minds of AI researchers. In this position paper, we lay out a small set of arguments that call into question a particular family of these assumptions. These assumptions all arise from the dominant paradigm of analytic materialist philosophy in combination with modern cognitive science: 1. consciousness is a property that emerges from complex matter, 2. consciousness depends on architectural cognitive complexity, and 3. consciousness can be measured as a function of behavioral cognitive complexity. Alternatives to these views and their implications for future machine consciousness research are discussed.

ABSTRACT Planaria exhibit remarkable regenerative ability, including the capacity to regrow complete heads and brains after decapitation. Here, we re-investigated whether regenerated planaria can preserve learned avoidance behavior, a phenomenon that has been reported previously but has been difficult to study due to unreliable experimental protocols. Using a light-to-food associative conditioning paradigm, planaria were trained to override their normal photophobic preference and then decapitated. Following a two-week regeneration period, behavioral responses to the conditioned stimulus were re-evaluated. Results indicated that the majority of regenerated planaria retained the learned response, supporting a model in which behavioral patterns can regenerate as well as anatomical patterns. By establishing a consistent, low-cost, and effective protocol for studying memory persistence through regeneration, such work may help inform future research on memory loss, resilience, and recovery in neurodegenerative diseases.

Abstract Metabolic networks are typically viewed as homeostatic systems that stabilize flux, energy charge, redox balance, and metabolite availability under perturbation. However, it remains unclear whether the same feedback architectures that support metabolic robustness can also generate learning-like, experience-dependent adaptation. Here, we develop a coarse-grained dynamical model of mammalian energy metabolism to test whether prior perturbation can improve future metabolic responses. The model represents core glucose, glutamine, fatty acid, and oxidative phosphorylation pathways as coupled ordinary differential equations with Michaelis–Menten-type fluxes, product-inhibition feedback, adaptive enzyme-capacity regulation, and explicit ATP costs for enzyme adjustment. Rather than aiming to reproduce quantitative fluxes for a specific cell type, the framework is designed to expose how metabolic feedback, regulatory cost, repeated perturbation, and environmental variability interact. We use this model to ask whether adaptive enzyme regulation enables improved recovery after repeated challenges, whether such effects depend on energetic control costs, and whether environmental variability broadens or constrains the set of reachable adaptive states. This approach provides a tractable way to investigate how homeostatic metabolic regulation may give rise to experience-dependent metabolic plasticity.

ABSTRACT Diffusion Models (DMs) and Evolutionary Algorithms (EAs) share a core generative principle: iterative refinement of random initial distributions to produce high‐quality solutions. DMs degrade and restore data using Gaussian noise, enabling versatile generation, while EAs optimize numerical parameters through biologically inspired heuristics. Our research integrates these frameworks, employing deep learning‐based DMs to enhance EAs across diverse domains. By iteratively refining DMs with heuristically curated databases, we generate better‐adapted offspring parameters, achieving efficient convergence toward high‐fitness solutions while preserving explorative diversity. DMs augment EAs with deep memory, retaining historical data and exploiting subtle correlations for refined sampling. Classifier‐free guidance further enables precise control over evolutionary dynamics, targeting specific genotypical, phenotypical, or population traits. This hybrid approach transforms EAs into adaptive, memory‐enhanced frameworks, offering unprecedented flexibility, and precision in evolutionary optimization, with broad implications for generative modeling and heuristic search.