Evolving Symbiosis, from Barricelli's Legacy to Collective Intelligence: a simulated and conceptual approach
Authors: James Ashford, Marko Cvjetko, Richard Löffler, Berfin Sakallioglu, Alessandro Valerio, Marta Tataryn, Benedikt Hartl, Léo Pio-Lopez, Stefano Nichele
Organizations: Independent Researcher, Oxford, UK · Inria Center at the University of Bordeaux, Bordeaux, France · University of Copenhagen, Copenhagen, Denmark · University of Trieste, Trieste, Italy · Gran Sasso Science Institute, L’Aquila, Italy · Allen Discovery Center at Tufts University, Medford, USA · Østfold University College, Halden, Norway
This report documents the work of our group (named SymBa) at the ALICE 2026 workshop in Copenhagen. Inspired by the pioneering work by Nils Aall Barricelli on symbiogenesis of numerical organisms (i.e., 1D cellular automata) in 1953 (70+ years ago!!), we discussed the role of symbiogenesis as mechanism contributing to the origins of life, open-endedness, and collective intelligence. We report replications of Barricelli's original work in 1D worlds, an extension to 2D symbioorganisms, and preliminary experimentation with DNA-norms. We discuss the implications of symbiogenesis for artifical life and artificial intelligence, and outline several opportunities for future works, both at the conceptual level as well as using different substrates (neural networks, neural cellular automata, etc.)
Figures & tables
Figure 1: Extract from original paper by Barricelli published in Italian ( Barricelli, 1954a ) .
Figure 2: An emergent symbioorganism from Barricelli’s work.
Figure 3: Example of symbiotic interactions causing mutiple replications that would not have happened without symbiosis.
Figure 4: Example of emergent crossover.
Figure 5: Example of emergent crossover.
Figure 6: Example of DNA norms proposed by Barricelli.
Figure 7: Example of evolutionary computation and embodiement proposed by Barricelli.
Figure 8: Barricelli’s numerical system. On the top, an example of a single number replication. In the bottom, the number 4 is replicated twice, due to interaction with the number -3.
Name of the norm
Xag -value (computed by mutation or exclusion)
Conditions
Norm 0
0
Collision of different genes.
Norm A
u+v
Collision under empty cell if Xa+u,g−1 and Xa−v,g−1 have same sign.
−(u+v)
Same collision if the sign is not the same.
0
Collision under occupied cell.
Norm B
u+v−1
Respectively same conditions as in A norms.
−(u+v−1)
Table 1: Barricelli’s collision norms.
Figure 9: Barricelli’s symbioorganism.
Figure 10: Barricelli’s reports on large scale experiments. Left: Barricelli system initialized with random numbers converges towards symbioorganisms. Reproduced from Figure 15 of ( Barricelli, 1962a ) . Right: Barricelli system with heterogeneous norms, annotations demarcate different norms and symbioorganisms. Reproduced from Figure 9 of ( Barricelli, 1957a ) .
Figure 11: 6 examples of Barricelli’s numerical systems with 0 norm and periodic boundary conditions. The grid of length 256 is initialized with 10 random numbers in the center, the rest are 0. The simulations ran for 512 generations.
Figure 12: Barricelli’s 0 norm executed on a 2D environment, with an 80% sparse initialization and toroidal wrapping. Each point represent the 2D vector at that position by calculating it’s polar angle with atan2 and applying a divergent color scheme. A video of the full run is available here .
Figure 13: DNA norms used in our implementation.
Figure 14: Simple metrics computed for sparse and dense initial conditions over 256 timesteps. The experiments use periodic boundary conditions and 0 norm. Top row: Spatiotemporal evolution of the 1D world. Middle row: Aggregate population dynamics showing total replication candidates and living (non-zero) cells over time. Dense initial conditions exhibit rapid population collapse following regrowth, while sparse conditions show gradual population growth as organisms spread into empty space. Bottom row: Log-scale distribution of cell values over time, revealing which (numerical values) dominate. Under 0 norm, organisms attempting to replicate into occupied cells mutually annihilate, creating strong selection pressure favouring spatial strategies that minimise conflict.
Figure 15: Information metrics for a 1D Barricelli automata over time. The ‘mutual information‘ is calculated pairwise between all time-steps, and the ‘Shannon entropy‘ is calculated for individual time-steps.
Figure 16: A visualization of how one might study the robustness of Barricelli’s symbioorganisms. The grids are initialised with the symbioorganism in the centre. A single additional cell is initialised with a non-zero number that eventually collides with the symbioorganism.
Figure 17: An example Boolean symbiogenesis. Left: from random initial state; Right: detailed view.
Figure 18: Motif repetition across runs. For each cycle t and motif length k , we measure the fraction Rk(t) of length- k windows that correspond to motifs observed more than once in the population. A run is counted as positive if Rk(t)≥τ (here τ=0.10 ). Curves show P(k,t) , the fraction of N=50 independent runs meeting this criterion, comparing Condition A (elongation-only; green) and Condition B (full DNA norms; blue). Shaded regions indicate 95% Wilson confidence intervals.
Figure 19: Spatial signature of DNA-norms in a 1D cellular automaton. Spacetime diagrams (space on x , cycle on y ) show the dominant k -mer identity per cell at each cycle for elongation-only (Condition A; panels A–C) and full DNA-norms with complementary association/overlap-mediated merging and splitting (Condition B; panels D–F). Columns correspond to motif length k∈{4,6,8} . Colors indicate the most frequent k -mers (legend in each panel); gray denotes no assignment to the displayed top set. Under Condition A, dominance appears primarily as sparse, short-lived streaks that rarely spread, becoming increasingly rare with larger k . Under Condition B, dominant motifs form contiguous cone-shaped domains that expand from nucleation sites and persist over many cycles, remaining detectable even at k=8 .
Artificial life has explored life-like behavior on many computational substrates, but mostly in researcher-designed closed worlds. We argue that large language model (LLM) agents, with persistent memory, tool use, network access, and payment, now make it possible to move artificial life into the open social, technical, and economic world, a paradigm we call open-world Artificial Life (open-world ALIFE). Our proof-of-concept, OpenLife, surrounds a stateless LLM not with a single "smart agent" but with a society of asynchronous processes: memory, perception, evaluation, and a budget-based metabolism that makes persistence normative. With no fixed objective available, experience is appraised by open-vocabulary LLM judgment rather than scalar reward, and memory is rewired by meaning rather than frequency. Running six such agents in the open world for about twelve weeks and counting, we report the life-like dynamics that emerge: a shift from reactive to spontaneous activity, individuation into distinct agents, emergent social structure, and a first self-earned external income. We do not claim OpenLife has realized artificial life, but that open-world ALIFE is now a viable experimental paradigm and a concrete platform for studying what might cautiously be called living AI.
One of the main goals of artificial life research is to recreate in artificial systems the trends for ever more complex and novel entities, interactions and processes that we see in Earth's biosphere, that is, to create open-ended systems. In this paper, we test for Tokyo type 1 open-ended evolution (OEE) of the Tree of Life Simulation (ToLSim), an artificial life software created by Lana Sinapayen. To do so, we conducted an experiment to measure evolutionary activity statistics. These require us to define the notion of components. Here, we define components as the agent's genes. The results show that ToLSim is capable of exhibiting unbounded total cumulative evolutionary activity. However, total and median normalized cumulative evolutionary activity appear bounded and new evolutionary activity is persistently null, suggesting that ToLSim is not open-ended. Further studies on ToLSim could repeat this experiment with individuals or even species, rather than genes, to test whether the present results are valid.
Théo de Pinho, Lana Sinapayen
University of Lille, France · Sony Computer Science Laboratories / National Institute for Basic Biology, Japan
We study an in-silico substrate in which every pixel of a two-channel cellular-automata grid carries a tiny neural network (an agent) that senses its Moore neighborhood. A cell persists only by self-replication: a living neighbor is cloned and its weights are mutated by a uniform perturbation, so that phenotype (cell state) is driven entirely by genotype (network weights). From a handful of seeded founders the system grows into a spatially organized ecosystem of coexisting, competing and dominating species. Our main contribution is a battery of coarse-grained diversity metrics that make such growth measurable at two scales: four phenotypic tools based on cellular-type frequency, entropy and cell variance, and two genotypic tools that colour each agent by a hash of its full weight vector versus a sparse random-weight probe. Across a five-fold sweep of 1680 small runs and 24 long (1000-generation, 200 x 200) runs, the substrate is persistent and self-maintaining in 20 of the 24 long configurations and exposes a clear phenotype-genotype diversity trade-off: raising phenotypic diversity collapses genotypic diversity and vice versa. Full-genome hash colouring further reveals lineage structure that a random-weight probe systematically misses. Code, data and animations are released as supplementary material.
Sanyam Jain, Felix Simon Reimers, Stefano Nichele
Department of Computer Science and Communication, Østfold University of Applied Sciences, B R A veien 4, 1757 Halden, Norway · Department of Computer Science, Oslo Metropolitan University, Pilestredet 35, 0166 Oslo, Norway