The shutdown problem is the problem of programming an agent so that it behaves useful during normal operation and facilitates a shutdown if and only if the creator wants to shut the agent down. First, we revisit a formalisation of this problem from the literature and we show that solutions are essentially unique. Second, we formally define ad-hoc constructions. Last, we present one trivial ad-hoc construction for the shutdown problem and show that every solution to the shutdown problem must come from an ad-hoc construction, which is to be expected given the uniqueness from the first point. We relate this to non-existence theorems from the literature.

Several propagating and stationary patterns have been observed in the model of excitable media on face-centered cubic lattice. In this research, we consider a propagating pattern called type-I glider on the lattice as a flying bit 0. By colliding it with another propagating pattern called type-II glider in the proper configuration, it can be turned at right angles. Colliding a type-I glider with another in the other proper configuration converts it into a type-IV glider which represents a flying bit 1. Furthermore, the collision of a flying bit 1 with a stationary pattern can convert the flying bit 1 into 0. These results lead to the possibility of creating logic gates in excitable media.

This project aims to create an affordable macroscopic physical experiment using simple principles to explore pattern formation and dynamics. Combining the Cheerios effect, a wellobserved phenomenon in fluid dynamics, with the geometric concept of aperiodic monotiles makes it possible to observe the self-assembly of complex structures from identical elements. Aperiodic monotiles are unique geometric shapes with a notable property: they can tile an infinite plane without forming a repeating pattern. The specific geometric properties of the monotiles influence the resulting formations. Perturbations can increase the complexity of clusters and make them evolve and interact with each other. This setup facilitates the self-organization of patterns on the liquid surface.

The regulation of growth, aging, and lifespan share fundamental genetic and metabolic pathways. An unsolved problem for growth regulation is understanding the mechanisms through which cells, or groups of cells, coordinate when to initiate a new phase of growth, such as puberty. Here, we propose a novel mechanism called decentralized cellular time-keeping as a method by which an individual cell, as part of a network of cells, can estimate the age of the network. Decentralized timekeeping is based on a cell mapping increasing information entropy in intercellular communication to organism age. We demonstrate with a computational simulation how such a system could regulate growth.

The evolution of controllers for heterogeneous Multi-Agent Systems, wherein two or more agents interact, is a fascinating area of research. While using evolutionary algorithms is a straightforward approach, the resulting behavior strongly depends on the capabilities of each individual and the type of established interaction. Recently, a novel technique, called n-mates evaluation, has been proposed to better estimate the contribution of individuals and generate more adaptive agents. However, the significance of parameter n has not been investigated. In this work, we analyze the impact of the parameter on the performance and adaptability of two agents collaborating to solve various evolutionary benchmark problems (i.e., foraging, escaping, aggregation). Statistical analysis demonstrates that selecting n = 5 results in superior outcomes across all investigated parameters ( n ∈ [1, 2, 5, 10]).

This work examines chemical autonomous agents — minimal chemical reaction networks exhibiting dissipation, autocatalysis, homeostasis, and associative learning — under the lens of the free-energy principle — a normative account of adaptive systems. The free-energy principle allows us to 1) identify the partition of system states belonging to the agent, 2) uncover how that agent synchronises its internal states to its environment, 3) understand the agent’s environmental fitness from the reaction rate constants. This initial work suggests that the free energy principle can provide a systematic approach to decompose, analyse, and understand complex adaptive systems and artificial life.

Living organisms need to solve rooting problems by selecting sets of paths linking several points of interests. Path selection processes are ubiquitous and ecologically valuable in plants, fungi, ameboids, and in animals like central-place foragers. However, path selection can be difficult to study, as some selection cues are not always visible (e.g. pheromones in insects). In this paper we present a method to study dynamical path selection behaviour solely from time-lapse images. We demonstrate how it can be applied to study freely roaming termites behaviour. Our method reconstructs the network of all paths ever taken and quantifies the amount of individuals for each edge at each time-step. The path selection behaviour can then be studied with null models. One can test for individual rules by simulating agents or running differential equations models in the network. Behavioural hypotheses can then be tested by comparing with observed networks’ properties, at any network scale and any time-step. We used this method to reconstruct freely roaming termites networks. We used differential equations to model the diffusion of termites in the network solely based on arbitrary turning preferences. Surprisingly, the common feature of animals to follow borders (thigmotaxy) emerged from this simple rule. However, our model lacked an amplifying mechanism to reproduce the intensity of termites’ path selection process. This was expected, as pheromones were not implemented in the model. This method could be used on any time-lapse data of different species, and authors are open for collaboration.

The sustained growth of a population of protocells which undergo symmetrical division (where each individual splits into two equal daughter protocells) requires synchronization between the two processes of (i) duplication of the genetic material and (ii) fission of the lipid container. It has however been observed that one often encounters uneven division, where daughters of different sizes may be generated. Here we analyze the case of asymmetrical division, where each protocell has exactly two daughters of different sizes. In this case no true synchronization is possible, and we introduce the notion of homogeneous growth which guarantees that sustained population growth is possible. We consider different abstract models of protocells growth and reproduction and we show by simulation that homogeneous growth is encountered, both in Surface Reaction Models, where the replicators are located in the membrane, and in Internal Reaction Models where they are found in the internal water phase, under a broad set of different kinetic equations. We argue that, when there are different kinds of replicators, it is legitimate to identify the “chemical signature” of the protocell with the set of the ratios between the quantities of these replicators at fission time: it is shown that, in the case of linear kinetic equations, the ratios and therefore the chemical identity are conserved through generations.

Neural cellular automata represent an evolution of the traditional cellular automata model, enhanced by the integration of a deep learning-based transition function. This shift from a manual to a data-driven approach significantly increases the adaptability of these models, enabling their application in diverse domains, including content generation and artificial life. However, their widespread application has been hampered by significant computational requirements. In this work, we introduce the Latent Neural Cellular Automata (LNCA) model, a novel architecture designed to address the resource limitations of neural cellular automata. Our approach shifts the computation from the conventional input space to a specially designed latent space, relying on a pre-trained autoencoder. We apply our model in the context of image restoration, which aims to reconstruct high-quality images from their degraded versions. This modification not only reduces the model’s resource consumption but also maintains a flexible framework suitable for various applications. Our model achieves a significant reduction in computational requirements while maintaining high reconstruction fidelity. This increase in efficiency allows for inputs up to 16 times larger than current state-of-the-art neural cellular automata models, using the same resources.

DNA nanotechnology has introduced the ability to create structures at the molecular scale, which is a promising approach for the implementation of very large swarms. However, the movement of such structures is heavily influenced by their size, prompting shape design optimization. Here, we use a quality-diversity approach to optimize the size of structures assembled from sets of DNA strands. We introduced a surrogate model to accelerate evaluations, with the ground truth provided by oxDNA, a physics-based simulator. We then iterate between optimization rounds using the QD algorithm, direct evaluation of promising and potentially mispredicted sets with oxDNA, and training of the surrogate model. We show that this approach efficiently generates diverse candidate sets at a fraction of simulation costs. Additionally, the surrogate model is reusable, enhancing the overall performance of future optimization tasks.

We consider various setups where large language models (LLMs) communicate solely with themselves or other LLMs. In accordance with similar results known for program representations (like λ-expressions or automata), we observe a natural tendency for the evolution of self-replicating text pieces, i.e., LLM prompts that cause any receiving LLM to produce a response similar to the original prompt. We argue that the study of these self-replicating patterns, which exist in natural language and across different types of LLMs, may have important implications on artificial intelligence, cultural studies, and related fields.

This paper introduces a Gene Regulatory Neural Cellular Automata (ENIGMA), an innovative extension of the Neural Cellular Automata (NCA) framework aimed at modeling biological development with a greater degree of biological fidelity. Traditional NCAs, while capable of generating complex patterns through neural network-driven update rules, lack mechanisms that closely mimic biological processes such as cell-cell signaling and gene regulatory networks (GRNs). Our ENIGMA model addresses these limitations by incorporating update rules based on a simulated gene regulatory network driven by cell-cell signaling, optimized both through backpropagation and genetic algorithms. We demonstrate the structure and functionality of ENIGMA through various experiments, comparing its performance and properties with those of natural organisms. Our findings reveal that ENIGMA can successfully simulate complex cellular networks and exhibit phenomena such as homeotic transformations, pattern maintenance in variable tissue sizes, and the formation of simple regulatory motifs akin to those observed in developmental biology. The introduction of ENIGMA represents a significant step towards bridging the gap between computational models and the intricacies of biological development, offering a versatile tool for exploring developmental and evolutionary questions with profound implications for understanding gene regulation, pattern formation, and the emergent behavior of complex systems.

Neurons are morphologically diverse, but the evolutionary advantage of this is unclear. In addition, neurons spike and exploit time in their computations, outputs and learning. However, most work on artificial neural networks (ANNs) abstract over these details and restrict learning and adaptation to the spatial parameters of weights and biases. Even when time is introduced, it is introduced through recurrency at a fixed time step (synchronous computation), and again, learning is restricted to weights and biases. Here we adapt weights, time constants and delays in an evolutionary context in an attempt to gain some insights into why neurons are so diverse. We show that nature might have evolved a morphologically diverse set of neurons to i) map spatio-temporal spike trains and ii) ease the evolutionary search for high performing solutions.

We introduce a simulation environment to facilitate research into emergent collective behaviour, with a focus on replicating the dynamics of ant colonies. By leveraging real-world data, the environment simulates a target ant trail that a controllable agent must learn to replicate, using sensory data observed by the target ant. This work aims to contribute to the neuroevolution of models for collective behaviour, focusing on evolving neural architectures that encode domain-specific behaviours in the network topology. By evolving models that can be modified and studied in a controlled environment, we can uncover the necessary conditions required for collective behaviours to emerge. We hope this environment will be useful to those studying the role of interactions in emergent behaviour within collective systems.

Neural Cellular Automata (NCA) models are trainable variations of traditional Cellular Automata (CA). Emergent motion in the patterns created by NCA has been successfully applied to synthesize dynamic textures. However, the conditions required for an NCA to display dynamic patterns remain unexplored. Here, we investigate the relationship between the NCA architecture and the emergent dynamics of the trained models. Specifically, we vary the number of channels in the cell state and the number of hidden neurons in the MultiLayer Perceptron (MLP), and draw a relationship between the combination of these two variables and the motion strength between successive frames. Our analysis reveals that the disparity and proportionality between these two variables have a strong correlation with the emergent dynamics in the NCA output. We thus propose a design principle for creating dynamic NCA.