The Convergence of Neuroscience and Computing: A Comprehensive Analysis of Wetware Computers and Organoid Intelligence

Full report

On this page

Introduction: The Post-Moore’s Law Landscape and the Biological Imperative

The trajectory of modern computing has been largely defined by the relentless miniaturization of silicon-based transistors, a paradigm famously encapsulated by Moore's Law. However, as the semiconductor industry approaches the fundamental physical limitations of atomic-scale fabrication, the energy requirements and thermal constraints of conventional hardware have begun to pose insurmountable barriers to the continued scaling of artificial intelligence (AI)1. The training and deployment of large-scale artificial neural networks (ANNs), which require vast arrays of graphics processing units (GPUs) and application-specific integrated circuits (ASICs), demand immense computational resources. This insatiable need for processing power contributes to significant energy consumption and environmental impact, pushing traditional computing architectures to their breaking point. In response, an interdisciplinary synthesis of neurobiology, stem cell engineering, materials science, and computer science has birthed a novel computational paradigm: wetware computing, or organoid intelligence (OI)3. Wetware computing represents a radical departure from traditional silicon-based architectures. Rather than simulating biological neural networks via digital abstractions in software, wetware computing utilizes living neural tissue—specifically, in vitro brain organoids or cultured biological neural networks (BNNs)—as the physical substrate for computation1. This field, often referred to as biological computing or synthetic biological intelligence, harnesses the intrinsic electrophysiological properties, neuroplasticity, and self-organizing capabilities of living neurons to perform complex data processing tasks6. The concept of wetware computing traces its conceptual origins to early cybernetics and cyberpunk fiction, but it transitioned into empirical science in the late 1990s when biological engineer William Ditto constructed a rudimentary neurocomputer capable of basic arithmetic using leech neurons1. Today, the field has evolved exponentially due to breakthroughs in human induced pluripotent stem cell (iPSC) technology, high-density microelectrode arrays (HD-MEAs), and microfluidic life-support systems8. By cultivating three-dimensional, miniature brain-like structures in the laboratory, researchers can now interface living human neural networks with digital inputs and outputs, creating closed-loop biohybrid systems capable of sensory processing, simulated motor control, and unsupervised learning11. The profound potential of this emerging domain was recently underscored by the United States National Science Foundation (NSF), which invested 14 million dollars into the Emerging Frontiers in Research and Innovation (EFRI) Biocomputing through EnGINeering Organoid Intelligence program to foster fundamental, ethically responsible research14. The implications of this technology extend far beyond the creation of energy-efficient coprocessors. Organoid intelligence promises to revolutionize our understanding of the human brain, offering unprecedented in vitro models for studying neurodevelopmental processes, memory consolidation, and the pathophysiology of neurodegenerative diseases6. Furthermore, as these biological systems demonstrate emergent capabilities akin to rudimentary cognition—such as learning to play video games or recognizing speech patterns—they catalyze profound ethical and philosophical debates regarding consciousness, moral status, and the future relationship between humanity and bio-engineered intelligence16.

The Architectural Dichotomy: Silicon Substrates Versus Biological Neural Networks

To understand the transformative potential of wetware computing, one must first examine the fundamental differences between how silicon-based architectures and biological brains process information. Modern computers operate on the von Neumann architecture, which strictly separates data storage (memory) from data processing (the central processing unit)2. This physical separation necessitates the constant shuttling of data back and forth across a system bus, creating the infamous "von Neumann bottleneck." This bottleneck severely throttles processing speed and dramatically increases the energy consumption required to move data, especially in data-intensive applications like machine learning2. In stark contrast, the human brain fuses memory and computation seamlessly within its network of 86 billion neurons and hundreds of trillions of synapses2. Every neuron functions simultaneously as a processor and a memory storage unit, communicating via highly parallel, asynchronous electrochemical impulses3. This structural divergence results in an astonishing disparity in energy efficiency.

The energy efficiency of biological tissue is perhaps its most compelling computational attribute. A biological neuron is estimated to be one million times more energy-efficient than an artificial neuron simulated on digital hardware20. While the Oak Ridge National Laboratory's Frontier supercomputer requires 21 megawatts of power and massive cooling infrastructure to achieve 1.1 exaflops of processing power, the human brain achieves a comparable computational throughput using a mere 20 watts5. If scaled appropriately, brain-based biocomputers could theoretically meet the data storage capacity of the entire United States using only 1,600 kilowatts of energy, fundamentally altering the economics of data centers9. Beyond energy dynamics, biological neural networks exhibit fundamentally different learning mechanisms. Conventional AI algorithms, particularly deep learning models, require millions of training iterations and colossal datasets to recognize patterns9. For example, the AlphaGo system required training on 160,000 historical games to achieve mastery in the game of Go9. In contrast, biological brains exhibit exceptional "sample efficiency." A human can learn to distinguish a cat from a dog after viewing fewer than ten examples9. This biological learning (BL) relies on dynamic network plasticity—the ability of neurons to continuously alter their chemical conformations, forge new synaptic connections, and prune unused pathways in real-time1. Biological systems also do not suffer from "catastrophic forgetting," a common flaw in artificial neural networks where training a model on a new task frequently overwrites its competence in a previous one21. Furthermore, biological systems exhibit a profound capacity to process noisy, incomplete, and highly heterogeneous data4. Individual neurons do not operate in simple binary states; instead, they shift through thousands of dynamic configurations, redirecting electrical pulses through hundreds of thousands of ion channels across vast synaptic webs1. This continuous, nonlinear dynamic behavior provides biological hardware with a processing richness that current digital emulation struggles to replicate.

Theoretical Frameworks Governing Biological Computation

The self-organizing behavior of biological neural networks is frequently analyzed through theoretical frameworks developed in computational neuroscience, most notably the "free energy principle" pioneered by theoretical neuroscientist Karl Friston11. The free energy principle posits that all biological systems are driven by an existential imperative to minimize the unpredictability (or entropy) of their sensory environments21. In a computational context, a cultured biological neural network will spontaneously adapt its internal synaptic weights and firing patterns to render its environment more predictable11. When living neurons are embodied in a simulated environment and provided with structured electrophysiological feedback, they adjust their activity to avoid chaotic or random stimuli21. This innate drive toward environmental predictability allows researchers to establish closed-loop reinforcement learning paradigms without the need for traditional, mathematically rigid backpropagation algorithms. By providing predictable, structured electrical signals (such as a specific frequency pulse) as a "reward" and chaotic, unpredictable noise as a "punishment," a biological network can be compelled to alter its behavior, effectively learning to perform targeted computational tasks17. This emergent goal-directed behavior forms the theoretical bedrock for practical wetware computing implementations. It suggests that neural correlates of learning—such as the coexistence of Pearce-Hall and Rescorla-Wagner associative learning mechanisms deep within the brain—can be abstracted and applied to in vitro cell cultures24. The resulting biological computation allows these living architectures to continuously adapt to new stimuli, optimizing their energy output while simultaneously learning to process external inputs.

Reservoir Computing: Bridging the Digital and Biological Divide

To harness the chaotic, high-dimensional activity of living neurons for practical computation, the field of wetware computing has heavily adopted a machine learning paradigm known as "reservoir computing" (RC)2. Reservoir computing is a framework derived from recurrent neural network (RNN) theory that radically simplifies the training of complex, non-linear systems25. In conventional reservoir computing, an input signal is projected into a "reservoir"—a fixed, highly non-linear, and recurrently connected dynamic system25. The reservoir maps the lower-dimensional input into a much higher-dimensional computational space. Crucially, the internal synaptic weights of the reservoir itself are not explicitly trained or mathematically updated by the programmer; the reservoir is treated entirely as a "black box"25. Instead, only a simple, linear readout layer (or decoder) at the output is trained to interpret the complex states of the reservoir and map them to the desired classification or prediction, minimizing a loss function usually obtained via simple linear regression25. Historically, researchers have utilized various physical systems as reservoirs. These range from literal buckets of water (where the rippling surface acts as the non-linear transformation medium) to complex electronic systems like memristors, 2-D quantum dot lattices, and the Gaussian states of interacting quantum harmonic oscillators25. However, Biological Reservoir Computing (BRC) replaces these inanimate physical systems with living neural networks cultured in vitro12. Biological neural networks are uniquely suited for reservoir computing due to their intrinsic properties: massive connectivity, global synchrony, nonlinear signal integration, and fading memory (the ability to retain traces of recent inputs, which is essential for processing time-series data)2. When a biological reservoir is stimulated via input electrodes, the chaotic, spontaneous background activity of the neural culture transforms into structured, low-dimensional dynamics12. The varying spike trains generated by the neurons are captured by recording electrodes and translated into feature vectors27. A standard linear regression algorithm or a logistic classifier is then applied to these feature vectors to produce the final computational output2. What elevates biological reservoirs above classical physical reservoirs (like quantum dots or memristors) is the biological substrate's inherent neuroplasticity. While traditional physical reservoirs maintain fixed internal dynamics throughout the computation, a brain organoid physically alters its synaptic connectivity over time in response to repeated stimulation2. This unsupervised learning and network reshaping within the biological reservoir significantly enhances the system's ability to classify temporal signals and perform complex feature engineering12.

The Bioengineering Infrastructure of Wetware Computers

The construction of a functional wetware computer is a multidisciplinary endeavor that requires a delicate convergence of tissue engineering, microelectronics, and microfluidics. A complete biological computing platform—often referred to as a "Brain-on-a-Chip"—comprises three primary components: the biological substrate (human brain organoids), the electrophysiological interface (microelectrode arrays), and the life-support infrastructure (microfluidic systems)8.

Stem Cell Technologies and the Cultivation of Cerebral Organoids

The biological hardware at the core of organoid intelligence is derived from human pluripotent stem cells (hPSCs), which encompass both embryonic stem cells (ESCs) and induced pluripotent stem cells (iPSCs)6. The utilization of iPSCs is particularly favored due to their ethical advantages and their potential for personalized medicine. iPSCs are generated by reprogramming adult somatic cells (such as skin fibroblasts or peripheral blood mononuclear cells) back into an embryonic-like pluripotent state, allowing researchers to grow brain tissue matching the exact genetic profile of a human donor9. Once the iPSCs are established, they are carefully directed through a multi-stage differentiation protocol to generate cerebral organoids. The most widely adopted methodologies iterate upon the groundbreaking protocols established by Madeline Lancaster31. The generation process requires precise biochemical orchestration, generally spanning several weeks and encompassing four distinct biological stages31:

A mature brain organoid typically measures around 500 micrometers to several millimeters in diameter and contains between 100,000 and several million cells5. Although they remain a fraction of the size of a human brain, these dense networks generate spontaneous electrical activity, synchronized local field potentials (LFPs), and oscillatory dynamics akin to those observed in human cortical tissue6.

The Electrophysiological Interface: High-Density Microelectrode Arrays (HD-MEAs)

To utilize an organoid for computation, researchers must establish a bidirectional communication channel capable of delivering input data and recording the network's processed output. This is achieved through the use of high-density multi-electrode arrays (HD-MEAs)2. The organoid is mounted directly onto the MEA surface, allowing the biological tissue to integrate physically with the silicon hardware13. The MEA contains hundreds or thousands of microscopic electrodes that serve dual functions. First, they act as stimulators, converting digital data into spatiotemporal patterns of electrical pulses that trigger action potentials within targeted neural clusters11. Current interfacing methods map specific arrays of electrodes to distinct input variables; for example, the physical location of an active electrode might dictate the source of a stimulus, while the frequency of the electrical pulse encodes the magnitude of that stimulus11. Second, the electrodes act as sensors, detecting the faint extracellular voltage changes generated by the firing neurons21. The resulting spike trains generated by the network are processed through complex spike detection and artifact removal algorithms before being passed to a digital decoder8. A significant challenge remains in managing the electrical noise generated by random neuronal firing and translation artifacts, which can obscure the true signal and result in variable computational outputs23.

Microfluidic Life-Support and Environmental Homeostasis

A fundamental challenge in wetware computing is keeping the biological processors alive, healthy, and stable for extended periods. Biological neurons require strict homeostasis, necessitating precise regulation of temperature, pH, osmolarity, and nutrient concentrations35. To address this, modern brain-on-a-chip platforms integrate sophisticated microfluidic networks5. Microfluidics involves the manipulation of fluids through microscopic channels. In a biological computer, computer-controlled perfusion pumps continuously drive oxygenated, nutrient-rich culture media across the organoid while simultaneously siphoning away toxic metabolic byproducts3. This continuous, automated flow eliminates the need for manual media changes, reducing the risk of contamination and minimizing physical disruptions that could alter the network's learned synaptic weights3. Research institutions such as the Tay Lab at the University of Chicago and Northwestern University have pioneered advanced microfluidic cell culture systems that enable highly multiplexed stimulation and high-throughput time-lapse imaging38. These systems utilize microfluidic combinatorics to create spatiotemporal gradients and dynamically alter ligand mixtures over time, supporting the scalable cultivation of thousands of dynamical cell cultures38. Such advanced microfluidic setups allow for the precise delivery of neurotransmitters, neuropharmacological drugs, or chemical signals to specific regions of the organoid. This capability serves as a secondary, chemical input modality for computation, significantly expanding the I/O bandwidth beyond pure electrical stimulation5. Furthermore, integrated optical systems utilizing UV lights allow for the localized uncaging of specific molecules, facilitating highly complex, closed-loop neuro-engineering studies3.

Neuromorphic and Biomimetic Paradigms in National Laboratories

The push toward biological and bio-inspired computing has garnered substantial investment from the defense and energy sectors, seeking solutions for extreme environments where traditional computing fails. The United States Department of Energy (DOE) national laboratories are deeply engaged in bridging the gap between solid-state neuromorphic chips and true biological wetware. At the Argonne National Laboratory, scientists have pioneered neuromorphic computer chips modeled explicitly after the brains of insects, such as bees and fruit flies39. Acknowledging that the human brain's vast number of neurons is difficult to replicate in near-term hardware, researchers analyzed the evolutionary efficiency of insects. Bees process complex navigation, sensory integration, and wind orientation while consuming merely milliwatts of power39. The Argonne team developed a nanocomposite material—tungsten aluminum oxide—to create neuromorphic chips capable of filtering connections and applying neural weights in real-time, functioning at power levels far below one watt39. These bio-inspired systems are actively being developed for deployment in extreme environments, such as smart city sensors detecting poisonous gases or autonomous monitors deployed inside hot nuclear reactors39. Concurrently, at the Los Alamos National Laboratory (LANL), computational neurologists are striving to scale neuromorphic architectures to human-brain levels of complexity. Current large neuromorphic computers house slightly over a billion artificial neurons connected by 100 billion synaptic connections41. However, LANL researchers project that a neuromorphic computer housing the same number of neurons as the human cerebral cortex could soon be contained within a two-square-meter box41. Such a machine would operate up to one million times faster than a biological brain while consuming roughly 10 kilowatts of power—comparable to a home air conditioning unit41. These advancements in solid-state neuromorphic computing serve as critical stepping stones, proving that highly scaled, parallel network topologies are fundamentally superior to von Neumann architectures, thereby validating the ultimate goal of transitioning to genuine biological wetware.

Empirical Milestones in Wetware Computation

While the theoretical concepts of organoid intelligence have existed for decades, recent empirical breakthroughs have conclusively demonstrated that in vitro neural networks can perform meaningful computational tasks. Three distinct projects currently lead the field in translating wetware theory into functional hardware.

Cortical Labs: DishBrain and the CL1 Platform

In 2022, researchers at the Melbourne-based biotechnology startup Cortical Labs published a landmark study in the journal Neuron, unveiling a system dubbed "DishBrain"11. The research team successfully cultured approximately 800,000 biological neurons—comprising both mouse embryonic cells and human stem cell-derived neurons—on a high-density MEA and integrated them with a simulated environment to play the classic video game Pong11. The digital game state was translated into electrical stimuli. Electrodes on the left or right of the array fired to indicate the spatial position of the virtual ball, while the frequency of the electrical pulses indicated the ball's distance from the paddle11. Leveraging the free energy principle, the system provided organized, predictable electrical feedback when the neurons successfully hit the ball, and chaotic, unpredictable noise when they missed11. Remarkably, within just five minutes of closed-loop gameplay, the neurons coordinated their spiking activity, spontaneously reorganizing their network to successfully track and return the ball21. This represented the first time that cultured biological neurons were stimulated in a structured manner to produce goal-directed, intelligent behavior11. Building upon DishBrain, Cortical Labs subsequently developed the CL1 system, a fully commercialized biological computer36. The CL1 integrates high-density MEAs under sterile microfluidic channels with automated perfusion pumps, allowing the "Silicon Neurons" to survive for months36. In early 2026, Cortical Labs showcased the CL1 controlling actions within the complex 1993 first-person shooter game DOOM36. The system mapped 48 stimulation channels to sensory input (game pixels translated into amplitude-modulated stimulation bursts) and 12 recording channels to game actions (move, turn, shoot commands interpreted by a temporal decoder)36. Utilizing an open-source Proximal Policy Optimization (PPO) reinforcement learning loop connected via WebSocket APIs, the network of 200,000 living neurons required nearly seven days of continuous training to exceed random shooting behavior36. Crucially, the entire hybrid stack consumed less than one watt of power36. This demonstration proved the immense viability of using biological tissue to navigate highly complex, three-dimensional simulated environments using a microscopic fraction of the power required by conventional GPUs.

Indiana University: The Brainoware System

A parallel milestone was achieved by a research team led by Feng Guo at Indiana University, who published their findings on the "Brainoware" system in Nature Electronics in 20232. Unlike DishBrain, which utilized a 2D monolayer of neurons, Brainoware mounted a fully developed, three-dimensional human brain organoid onto a multielectrode array2. Brainoware explicitly leveraged the biological reservoir computing framework. The researchers converted complex temporal data into spatiotemporal sequences of electrical stimulation and fed them into the organoid2. To test the system's pattern recognition capabilities, Brainoware was tasked with a speech recognition challenge. The researchers fed the organoid 240 audio recordings of eight different individuals speaking Japanese vowels18. Initially, the readout layer could only identify the specific speaker with 30-40% accuracy44. However, after two days of training (one epoch per half-day), the organoid underwent unsupervised learning, physically reshaping its functional connectivity to map the distinct audio features2. Consequently, the system's accuracy rapidly increased to approximately 78%, outperforming baseline metrics set by both simple logistic regression algorithms and standard digital echo state networks (ESN) under specific conditions18. Furthermore, Brainoware was tested on its ability to predict the Hénon map, a complex mathematical model illustrating nonlinear dynamic and chaotic behavior in mathematics13. The biological reservoir proved highly capable of processing the spatial information and fading memory required to solve the chaotic equations, demonstrating the system's robust capacity for advanced time-series analysis and nonlinear feature engineering28.

FinalSpark and the Remote Neuroplatform

While Cortical Labs and Indiana University have focused on proving the computational validity of living tissue, the Swiss biotechnology startup FinalSpark has focused on scalability, automation, and global accessibility3. Recognizing that the optimization of biological learning algorithms requires vast amounts of experimental data, FinalSpark launched the "Neuroplatform," the world's first remote API for biological computing3. The Neuroplatform hosts multiple brain organoids in highly automated incubators equipped with microfluidic systems that handle all medium flows and exchanges autonomously3. This automation reduces physical interventions, ensuring stable environmental conditions and allowing the organoids to survive and compute for over 100 days3. Over a three-year period, FinalSpark utilized over 1,000 brain organoids to generate more than 18 terabytes of electrophysiological data3. Through a dedicated Python library and Jupyter Notebook integration, researchers worldwide can lease time on the wetware hardware to conduct remote, 24/7 closed-loop experiments3. The API grants researchers granular control over the MEA's electrical stimulation, digital cameras, and microfluidic pumps3.

Translational Applications: Beyond Computation into Biomedicine

The maturation of organoid intelligence and wetware computing platforms yields benefits far beyond the realm of artificial intelligence. By combining biological computation with advanced analytics, these platforms are becoming the ultimate "biomimetic sandboxes" for translational medicine and neuropharmacology11. Historically, neurobiological research has relied heavily on animal models, which often fail to accurately replicate the complexities of the human brain, leading to notoriously high failure rates in clinical trials6. Two-dimensional cell cultures also fall short of representing the three-dimensional structural architecture and intercellular interactions found in vivo6. Brain-on-a-chip biocomputers bridge this gap by providing an active, learning human neural network whose cognitive and behavioral changes can be measured in real-time6. Researchers can generate brain organoids using iPSCs derived from patients suffering from neurodevelopmental disorders or neurodegenerative diseases such as Alzheimer's, Parkinson's, epilepsy, and dementia6. By observing how these genetically specific organoids behave during standardized computational tasks, scientists can identify the precise synaptic failures or neuroplasticity deficits driving the disease pathology9. Pharmaceutical companies can rapidly utilize these platforms for high-throughput drug screening, toxicity testing, and the development of personalized therapies7. By introducing therapeutic compounds into the microfluidic flow of an active wetware computer, researchers can observe directly how a drug alters neurotransmission, learning rates, and information processing at the synaptic level11. For instance, researchers at Cortical Labs have proposed experiments tracking the cognitive performance of a "drunken DishBrain" exposed to alcohol, providing granular data on exactly how intoxication degrades structural neural connectivity and adaptive behavior11. Similar software advancements, such as the OrganoID platform, automatically trace and quantify morphological changes in various organoids (including pancreatic cancer organoids) over time, allowing for rapid, large-scale phenotypic drug screening without human intervention46. Additionally, the integration of biological recording with AI has led to the creation of "digital twins" of brain organoids. By using the massive troves of electrophysiological data generated by platforms like FinalSpark, computational biologists can train advanced in silico models to predict how specific neural circuits will react to varied stimuli, bridging the gap between biological computation and digital simulation47.

Engineering Bottlenecks and Scalability Challenges

Despite the rapid progression of wetware computing, the field faces substantial bioengineering and computational hurdles that must be overcome before organoid intelligence can achieve widespread commercial integration.

Vascularization and Tissue Longevity

The most pressing biological limitation is the physical scaling of the organoids. In the human brain, a dense and intricate network of blood vessels ensures that every neuron is constantly supplied with oxygen and nutrients9. In vitro brain organoids lack this innate vasculature, relying entirely on the passive diffusion of nutrients from the surrounding media9. As a result, organoids face a strict size limit; if they grow beyond approximately 500 micrometers in diameter, passive diffusion fails to reach the innermost cells, leading to a necrotic core of dead tissue9. To build biological computers capable of rivaling advanced silicon systems, researchers must scale these networks to contain billions of neurons. This requires the development of artificial vasculature. Advanced bioengineering techniques, such as incorporating endothelial cells to encourage angiogenesis or 3D printing complex microfluidic vascular networks directly into the organoid scaffold, are currently under intense investigation9. Until a robust vascular solution is achieved, wetware computing will remain limited to relatively small neural clusters.

Interfacing Constraints and I/O Bottlenecks

The second major bottleneck lies in the input/output (I/O) interface. Current generation MEAs consist of rigid, two-dimensional planar arrays23. When a spherical, three-dimensional organoid is placed on a flat MEA, the electrodes only make contact with a tiny fraction of the cells on the organoid's basal surface8. Consequently, the vast majority of the network's internal processing remains a hidden "black box," severely limiting the resolution of both stimulation and recording capabilities23. To fully harness the 3D topology of brain organoids, new bio-interfacing technologies are required. Researchers are actively developing soft, flexible mesh electronics that can envelop the organoid, as well as 3D penetrating microelectrodes that can record from the organoid's inner core without causing excessive tissue damage5. Additionally, alternative I/O modalities are being explored, including optogenetics and the integration of retinal organoids to provide direct optical sensory input5.

Standardization and Reproducibility

Silicon chips are manufactured in cleanrooms to identical, nanometer-level specifications, ensuring perfectly reproducible behavior across billions of units. Biology, however, is inherently chaotic and highly variable. The differentiation of iPSCs into brain organoids is highly sensitive to microscopic environmental changes, leading to significant batch-to-batch heterogeneity23. Two organoids grown from the same stem cell line using the exact same protocol may develop drastically different morphological structures, cellular ratios, and baseline electrical activity profiles28. This biological variability complicates the standardization of biological hardware. If a neural network is to be integrated into commercial computing infrastructure, it must perform reliably and predictably. Scaling organoid intelligence will demand rigorous standardization of manufacturing protocols, highly automated robotic culture systems, and advanced quality control metrics to ensure consistent biological processor performance15.

Ethical, Philosophical, and Cyberbiosecurity Implications

As researchers successfully construct living networks that exhibit goal-directed behavior, learn from their environments, and process complex data, the field of organoid intelligence collides inevitably with profound ethical and philosophical quandaries. The creation of "intelligence-in-a-dish" necessitates the immediate development of robust neuroethical frameworks5. The primary ethical concern revolves around the potential emergence of sentience, consciousness, or the capacity to experience suffering (nociception) within these biological computers4. While the 200,000 neurons playing DOOM in a laboratory do not possess the structural complexity, sensory integration, or biological scale to harbor human-like consciousness, the line between an unfeeling biological processor and a sentient entity becomes increasingly blurred as researchers expand organoid size, interconnect multiple brain regions (creating complex "assembloids"), and equip them with continuous sensory data streams5. If a future wetware computer achieves a form of synthetic biological sentience, society will face unprecedented questions regarding the moral status, legal rights, and ethical treatment of biohybrid machines30. Another critical ethical dimension is the issue of donor consent and the commercialization of human tissue. The iPSCs used to generate brain organoids are ultimately derived from human donors10. As these biological processors transition into lucrative commercial computing products, the rights of the original cell donors must be fiercely protected30. Donors must be fully informed that their cellular material may be cultured into neural networks used to drive artificial intelligence, control robotics, or test pharmaceuticals. Without robust informed consent protocols and equitable frameworks for intellectual property, the commercialization of OI risks exploiting economically disadvantaged donors and eroding public trust in bioengineering4. The "Baltimore Declaration" formally introduced Organoid Intelligence as an emerging field requiring immediate ethical oversight to manage these complexities35. Leading institutions have championed an "embedded ethics" approach, placing philosophers, ethicists, and legal scholars directly within bio-computing laboratories to continuously monitor and guide the research as it evolves5. Finally, the advent of wetware introduces entirely new vectors in bio-cybersecurity and cyberbiosecurity. As biological components are integrated with digital networks and edge computing devices, these hybrid systems could be targeted by novel forms of malware4. Threat actors could theoretically design digital intrusions that manipulate biological learning processes, maliciously alter microfluidic nutrient flows, or weaponize the biological substrate to generate erratic computations49. The defense and intelligence sectors are already establishing "Red Teams" to scrutinize the strategic implications of autonomous systems driven by biological intelligence, highlighting the dual-use nature of this profoundly powerful technology21.

Conclusion

The emergence of wetware computing and organoid intelligence marks a watershed moment in the histories of both computer science and neurobiology. By utilizing human brain organoids as living computational substrates, researchers have bypassed the strict binary limitations of silicon, unlocking biological reservoir systems that compute with extreme energy efficiency, adapt through continuous neural plasticity, and process highly complex nonlinear information. The proofs of concept established by systems like DishBrain, Brainoware, and the Neuroplatform conclusively demonstrate that biological learning can be harnessed to execute concrete computational tasks—ranging from speech recognition to navigating simulated digital environments. However, the realization of commercially viable biohybrid computers requires overcoming formidable engineering obstacles. The field must develop scalable artificial vasculature to support larger organoids, pioneer three-dimensional flexible electrode arrays to resolve the current I/O bottlenecks, and standardize tissue manufacturing protocols to guarantee uniform performance across biological processors. More urgently, the scientific community must navigate the profound ethical labyrinth surrounding the use of human tissue for commercial AI hardware, ensuring that the march toward synthetic biological intelligence is governed by rigorous moral frameworks. Ultimately, the true promise of organoid intelligence is not merely the creation of a faster or more efficient computer, but the creation of a technological paradigm that harnesses the fundamental mechanics of biological cognition.

Works cited

1. Wetware computer \- Wikipedia, https://en.wikipedia.org/wiki/Wetware\_computer

2. Brain organoid reservoir computing for artificial intelligence \- Kinam Park, http://kinampark.com/T-Polymers/files/All%20References/Cai%202023%2C%20Brain%20organoid%20reservoir%20computing%20artificial%20intelligence.pdf

3. Open and remotely accessible Neuroplatform for research in wetware computing \- Frontiers, https://www.frontiersin.org/journals/artificial-intelligence/articles/10.3389/frai.2024.1376042/full

4. Organoid intelligence \- Wikipedia, https://en.wikipedia.org/wiki/Organoid\_intelligence

5. Organoid intelligence (OI): the new frontier in biocomputing and intelligence-in-a-dish, https://www.frontiersin.org/journals/science/articles/10.3389/fsci.2023.1017235/full

6. Brain Organoid Computing \- an Overview \- ResearchGate, https://www.researchgate.net/publication/390176420\_Brain\_Organoid\_Computing\_-\_an\_Overview

7. Organoid intelligence (OI) \- The ultimate functionality of a brain microphysiological system, https://pubmed.ncbi.nlm.nih.gov/37009773/

8. Advanced Brain‐on‐a‐Chip for Wetware Computing: A Review \- PMC, https://pmc.ncbi.nlm.nih.gov/articles/PMC12412539/

9. Powering up the next generation of biocomputers with brain organoids \- Frontiers, https://www.frontiersin.org/journals/science/article-hubs/organoid-intelligence-a-new-biocomputing-frontier/explainer

10. Induced Pluripotent Stem Cells and Organoids in Advancing Neuropathology Research and Therapies \- PMC, https://pmc.ncbi.nlm.nih.gov/articles/PMC11083827/

11. Human brain cells in a dish learn to play Pong | UCL News, https://www.ucl.ac.uk/news/2022/oct/human-brain-cells-dish-learn-play-pong

12. Online supervised learning of temporal patterns in biological neural networks under feedback control | PNAS, https://www.pnas.org/doi/10.1073/pnas.2521560123

13. IU bioengineers are building the intersection of brain organoids and AI, https://blogs.iu.edu/iuimpact/2023/12/15/human-brain-tissuebioengineers-are-building-the-intersection-of-organoids-and-ai/

14. NSF invests $14M in bioengineered systems and ethical biocomputing research, https://www.nsf.gov/news/nsf-invests-14m-bioengineered-systems-ethical-biocomputing

15. Organoid factory: The recent role of the human induced pluripotent stem cells (hiPSCs) in precision medicine \- PMC, https://pmc.ncbi.nlm.nih.gov/articles/PMC9869172/

16. Researchers are building computers that run on brain organoids — but have neglected a major ethical issue \- CuratedSci, https://sciencenews.strategian.com/public\_html/2026/07/27/researchers-are-building-computers-that-run-on-brain-organoids-but-have-neglected-a-major-ethical-issue/

17. Brain cells in a dish learn to play Pong \- Medicine, Nursing and Health Sciences, https://www.monash.edu/medicine/news/latest/2022-articles/brain-cells-in-a-dish-learn-to-play-pong

18. Cyborg brain in a dish can do maths and 'listen' to what you're saying \- Scimex, https://www.scimex.org/newsfeed/cyborg-brain-in-a-dish-can-do-math-and-listen-to-what-youre-saying

19. FinalSpark: Home, https://finalspark.com/

20. 'Wetware': Scientists use human mini-brains to power computers \- Canadian Affairs, https://www.canadianaffairs.news/2025/10/20/wetware-scientists-use-human-mini-brains-to-power-computers/

21. DishBrain: Are pong-playing neurons the future of AI? \- Microscopy Australia, https://micro.org.au/news/dishbrain-are-pong-playing-neurons-the-future-of-ai/

22. Biocomputing: the promise of biological computing \- Polytechnique Insights, https://www.polytechnique-insights.com/en/columns/science/biocomputing-the-promise-of-biological-computingbrains/

23. Brain Organoid Computing – an Overview \- arXiv, https://arxiv.org/html/2503.19770v2

24. Brain cells in a lab dish learn to... (NPR News) \- Behind the headlines \- NLM \- NCBI, https://www.ncbi.nlm.nih.gov/search/research-news/17391/

25. Reservoir computing \- Wikipedia, https://en.wikipedia.org/wiki/Reservoir\_computing

26. An introduction to reservoir computing \- arXiv, https://arxiv.org/html/2412.13212v1

27. Biological Reservoir Computing: Harnessing Living Neurons for AI \- ERCIM News, https://ercim-news.ercim.eu/en143/special/biological-reservoir-computing-harnessing-living-neurons-for-ai

28. Brain Organoid Computing for Artificial Intelligence \- bioRxiv, https://www.biorxiv.org/content/10.1101/2023.02.28.530502v1.full-text

29. Brain organoids-on-chip for neural diseases modeling: History, challenges and trends, https://pmc.ncbi.nlm.nih.gov/articles/PMC12594919/

30. Organoid Intelligence: Bridging the Gap between Biological and Artificial Neural Networks, https://www.scirp.org/journal/paperinformation?paperid=148102

31. Protocol for Neural Organoid Culture \- STEMCELL Technologies, https://www.stemcell.com/neural-organoid-culture.html

32. Brain organoid differentiation protocol using human pluripotent stem... \- ResearchGate, https://www.researchgate.net/figure/Brain-organoid-differentiation-protocol-using-human-pluripotent-stem-cells-hPSCs-A\_fig1\_358908717

33. Guidelines for Manufacturing and Application of Organoids: Brain \- PMC, https://pmc.ncbi.nlm.nih.gov/articles/PMC11170118/

34. Robust production of uniform human cerebral organoids from pluripotent stem cells, https://www.life-science-alliance.org/content/3/5/e202000707

35. Organoid Intelligence: Toward Biohybrid Architectures Beyond Silicon \- ELEKS, https://eleks.com/research/organoid-intelligence/

36. Biological AI Computing: Lab-Grown Neurons Beat DOOM \- AI CERTs News, https://www.aicerts.ai/news/biological-ai-computing-lab-grown-neurons-beat-doom/

37. Open and remotely accessible Neuroplatform for research in wetware computing \- PMC, https://pmc.ncbi.nlm.nih.gov/articles/PMC11097343/

38. Research \- Tay Lab \- The University of Chicago, https://taylab.uchicago.edu/research.html

39. Argonne Team Looks to Insect Brains as Models for Computer Chip Innovation, https://www.nextgov.com/artificial-intelligence/2019/10/argonne-team-looks-insect-brains-models-computer-chip-innovation/160686/

40. Argonne explores how ants, bees, and fruit flies can be the next big buzz in artificial intelligence, https://www.anl.gov/article/argonne-explores-how-ants-bees-and-fruit-flies-can-be-the-next-big-buzz-in-artificial-intelligence

41. Neuromorphic computing: the future of AI | LANL, https://www.lanl.gov/media/publications/1663/1269-neuromorphic-computing

42. The computer that runs on human neurons | Technology \- EL PAÍS English, https://english.elpais.com/technology/2025-04-05/the-computer-that-runs-on-human-neurons.html

43. Brain organoid reservoir computing for artificial intelligence | Request PDF \- ResearchGate, https://www.researchgate.net/publication/376413285\_Brain\_organoid\_reservoir\_computing\_for\_artificial\_intelligence

44. Human Brain Cells Linked With Computer Can Recognize Speech | Extremetech, https://www.extremetech.com/science/human-brain-cells-linked-with-computer-can-recognize-speech

45. Brain Organoid Computing for Artificial Intelligence \- bioRxiv, https://www.biorxiv.org/content/10.1101/2023.02.28.530502v1.full.pdf

46. A versatile deep learning platform for tracking and analysis of single-organoid dynamics \- Tay Lab, https://taylab.uchicago.edu/uploads/9/1/8/0/91804060/Organoid\_plos\_2022.pdf

47. Digital Twins of brain organoids: new research using FinalSpark Neuroplatform, https://finalspark.com/digital-twins-of-brain-organoids/

48. When Brain Cells Learned to Code. The emergence of Organoid Intelligence… | by Dr. Jerry A. Smith | Medium, https://medium.com/@jsmith0475/when-brain-cells-learned-to-code-e9e47151fbdf

49. Organoids, Biocybersecurity, and Cyberbiosecurity—A Light Exploration \- MDPI, https://www.mdpi.com/2674-1172/3/2/7