Organic Computation: AI and the Logical Power of Fungal Networks

Artificial Intelligence has a huge energy problem. The solution might not come from silicon supercomputers, but from nature. Researchers at Unconventional Compu

The insatiable energy hunger of Artificial Intelligence is pushing the semiconductor industry toward a physical and ecological breaking point. While global data centers consume resources comparable to those of entire nations to train increasingly mammoth language models, a radical alternative is making headway in unconventional computing laboratories. It is not about designing denser silicon, but about delegating computation to biology itself.

Organic computing based on fungi is no longer a fascinating bio-art experiment or a literary metaphor, but a rigorous field of engineering research, guided by pioneers such as Andrew Adamatzky and his team at the Unconventional Computing Laboratory of the University of the West of England. The underlying idea is astonishing: using networks of living fungal mycelium, interconnected with AI software, to solve complex logical problems.

In this in-depth piece for the Scenarios and Reflections column, we will explore the engineering of fungal computers. The thesis we will advance, supported by emerging scientific literature, is that organic computing is not a bizarre alternative destined to replace silicon, but an essential complementary paradigm. We are entering an era in which intelligence will not derive solely from the explicit programming of bits, but will emerge spontaneously from the growth, adaptation, and bioelectricity of living networks.

1. Beyond Metaphor: Pure Fungal Computing

To understand how an organism without a brain can process data, we must observe the underground world. Mycelium, the network of filaments (hyphae) that constitutes the vegetative apparatus of fungi, is an extraordinary communication infrastructure. As demonstrated in the foundational paper published in Interface Focus (Royal Society), information within the fungus travels in the form of spikes of electrical activity and chemical signals, much like the action potentials of human neurons.

"Pure fungal computing" uses this network as a physical substrate. Researchers encode Boolean logical values (0 and 1) in the presence or absence of these electrical discharges. Studies published on arXiv ("Logics in Fungal Mycelium Networks") and on PMC have demonstrated that the natural resistance and capacitance (RC model) of mycelium, combined with its morphological variations, allows for non-linear transformations of signals. This means that, by strategically placing electrodes in fungal colonies such as Aspergillus niger or Pleurotus ostreatus, it is possible to implement actual logic gates (AND, OR, NOT) and complex circuits. The logical output is decoded simply by measuring the electrical response of the living substrate at the edges of the colony.

2. Living Memristors and Sustainable Neuromorphic Hardware

If fungal logic gates represent the processor, researchers have also found a way to create memory. One of the bottlenecks of modern AI is the physical separation between computing and memory units (the von Neumann bottleneck). Neuromorphic hardware seeks to imitate the human brain by merging these two functions through "memristors" (resistors endowed with memory).

A groundbreaking study from Ohio State University, published in PLOS One, demonstrated the feasibility of fungal memristors using the common mycelium of the Shiitake mushroom (Lentinula edodes). Interfaced with gold and tin electrodes, these fungal networks can literally be "trained" to retain information. Organic memristors have demonstrated the ability to maintain their functionality at frequencies up to 5.85 kHz, with a remarkable accuracy of 90±1%. Even more incredible, Shiitake-based fungal computers can be preserved through dehydration cycles and show an intrinsic resistance to cosmic radiation. This detail opens unimaginable scenarios: scalable, zero-impact, and eco-compatible computing platforms, perfect for data processing in long-range aerospace missions.

3. Hybridization: Biomimetic AI Meets the Fungus

The quantum leap occurs when organic computing stops operating in isolation and is fused with Artificial Intelligence software. We are witnessing the birth of bio-electronic hybridization, in which living mycelium is interfaced with analog-to-digital converters (ADCs) and microcontrollers to read and guide the processing of an artificial neural network.

The most fascinating experiment in this domain is described in an essay deposited on Zenodo that presents the MyceliumTransformer (ironically nicknamed Large Language Mushroom). Researchers integrated spores of Psilocybe cubensis within an AI framework based on the Transformer architecture. In a fixed electronic lattice of 5×5 nodes, the mycelium was allowed to grow in a controlled environment. As the biological network matured, it formed conductive pathways that served as physical synaptic weights. By combining this living hardware with reinforcement learning algorithms (which rewarded or punished the fungus by altering nutrients based on the predictive accuracy of its responses), the hybrid system was able to solve canonical logical tasks (such as the XOR problem) with an efficiency unattainable by silicon hardware alone. The integrated mycelium proved essential for processing signals in an adaptive and decentralized manner.

4. Solving Geometric Problems Through Growth

The true superpower of mycelium is not speed, but its unparalleled ability to solve complex geometric optimization problems. As highlighted by popular science articles and reports from the Universitat Oberta de Catalunya, fungal networks excel at problems that would bring classical supercomputers to their knees, such as the famous "Traveling Salesman Problem" (finding the shortest route to connect a series of points without ever passing through the same one twice).

While a traditional algorithm must iteratively calculate (using brute-force or heuristics) thousands of variants consuming enormous amounts of energy, the fungal network solves the problem… by living. By feeding, the fungus naturally grows by tracing the thermodynamically most efficient spatial paths between different food sources. In other words, the organism computes the optimal solution through its own biological growth patterns. This approach, known as growth-based computation, allows for the generation of Voronoi diagrams, Delaunay triangulations, and complex proximity graphs, spending a microscopic fraction of the electricity required by a GPU.

Key Operational Takeaways (Takeaways for Engineers and Researchers)

  • Design Hybrid, Not Substitutive Architectures: The tech industry must not look at mycelium as a replacement for the central CPU. Fungi offer unparalleled energy efficiency and resilience, but present colossal limitations in terms of signal propagation speed and statistical reproducibility. The operational goal is to create hybrid coprocessors: use silicon for the speed of serial computation and fungal tissue for spatial optimization and low-consumption decision-making.
  • Redefine Sensory Processing (Sensor Fusion): Mycelium networks are intrinsically decentralized and have evolved to integrate chemical, thermal, and tactile inputs over vast areas. Future environmental sensors (for precision agriculture or forest monitoring) could use local fungal networks as natural onboard processors, sending to central AI servers only the data already chemically pre-processed by the soil.
  • Integrate Biological Uncertainty into AI Models: Software development for bio-electronic interfaces requires a paradigm shift. The code must be error-tolerant and adaptive, ready to handle fluctuations, cell death, or chemical drifts of the mycelium. AI must not give rigid orders, but act in "symbiosis" with the organism, interpreting its spontaneous reactions.

Conclusions: Cultivating Intelligence

Organic computing is forcing us to abandon the arrogance of algorithmic determinism. For decades, we conceived artificial intelligence as the triumph of mathematical order over natural chaos, building isolated and climate-controlled silicon cathedrals to perform billions of perfect and linear calculations. Today, faced with the thermal and energy unsustainability of this race, we are returning to knock on the doors of the undergrowth.

The hybridization between the human algorithm and millennial mycelium opens dizzying philosophical scenarios. We are building an ecosystem in which intelligence is a collaborative and distributed process, in which software code dialogues with organic ramifications in a language made of ions and extracellular potential variations.

This perspective forces us to overturn our approach to computer engineering and to confront a disarming question: if a network of fungi, refined by billions of years of blind evolution, can literally "grow" toward the optimal geometric solution of a logical problem consuming energy close to zero, does it still make sense to stubbornly design aseptic computers that must be explicitly programmed for every single task, or will the true future of computation no longer consist of writing intelligence, but of patiently learning to cultivate it?

Bibliographic References and Sources

Article by the Editorial Team of La Bussola dell'IA