Wetware computing: definition, how it works and examples

Wetware computing is computing carried out by living neural tissue, such as cultured neurons or brain organoids, connected to electronics. Electrodes deliver encoded electrical inputs to the cells and record their responses, and software translates that activity into outputs. The name contrasts “wet” living material with conventional hardware and software.

Key takeaways

• Wetware computing uses living neurons as the processing element, with silicon handling input, output and control.
• It is a branch of biocomputing and the practical basis of organoid intelligence.
• The core interface is the microelectrode array, a technology with roots in the early 1970s.
• Published demonstrations include closed-loop game play (DishBrain) and reservoir-based speech recognition (Brainoware).
• Main challenges are tissue lifespan, variability between samples, interface bandwidth, and the lack of a programming model.

What does wetware computing mean? #

Computer scientists divide systems into hardware and software. “Wetware” adds a third category: the living, water-based tissue of a nervous system. For a long time the word was mostly a metaphor for the human brain. Today it also names a research area in which neurons grown in the lab are used as part of a computing system.

A wetware system can use flat 2D cultures of dissociated neurons or 3D brain organoids, which have more cells, more cell types and more complex structure. The neurons may come from rodents or, increasingly, from human induced pluripotent stem cells. Jordan et al. used the term directly in the title of their 2024 Frontiers in Artificial Intelligence paper describing a remotely accessible “Neuroplatform for research in wetware computing”.

How does wetware computing work? #

A wetware computer is a stack of layers, each of which must work for the whole to function.

The tissue #

Neurons communicate with action potentials and change their connections through synaptic plasticity. In culture they form spontaneously active networks that often fire in coordinated network bursts. This intrinsic activity is both a resource (it shows the network is alive and connected) and a challenge (it adds noise to any computation).

The interface #

Cells grow on a chip covered in electrodes. Thomas et al. described an early miniature microelectrode array for cultured cells in Experimental Cell Research in 1972. Modern high-density microelectrode arrays carry thousands of electrodes built on CMOS chips, which lets software record from and stimulate many sites at once. Our post microelectrode arrays explained covers the hardware.

The software loop #

The computing happens in a loop:

1. Encode the task, such as a ball position in a game, into stimulation patterns.
2. Stimulate chosen electrodes.
3. Record voltage traces and detect spikes, often with spike sorting.
4. Decode the activity into an action or answer.
5. Feed back new stimulation that depends on the result, which is called closed-loop stimulation.

Kagan et al. reported in Neuron (2022) that cultured neurons embedded in such a loop, called DishBrain, improved at a simplified version of Pong. Cai et al. reported in Nature Electronics (2023) that a brain organoid could serve as the nonlinear core of a reservoir computing system for speech recognition. For how learning shows up in these setups, see how neurons learn in a dish.

A short history of wetware computing #

The idea of computing with cultured neurons is older than the current wave of organoid research. A rough timeline:

• 1972. Thomas et al. publish a miniature microelectrode array for recording cultured cells, the ancestor of today’s interfaces.
• 2001. DeMarse et al. describe a “neurally controlled animat” in Autonomous Robots, in which a culture of rat neurons on an electrode array drove a simulated body and received sensory feedback.
• 2013 onward. Brain organoid methods, such as the cerebral organoid protocol of Lancaster et al. in Nature, make it possible to grow 3D human neural tissue from stem cells.
• 2022. Kagan et al. report the DishBrain Pong experiment in Neuron.
• 2023 to 2024. Brainoware, the organoid intelligence roadmap, and remotely accessible platforms such as the Neuroplatform appear in the literature.

What has changed is less the core concept than the tools: human-derived tissue, denser electrodes, and far more capable software for real-time analysis.

Wetware vs hardware vs software #

The table shows the central trade-off. Wetware brings adaptive, self-organizing behavior that is hard to engineer, but it gives up the precision and permanence that make conventional computers dependable.

Why does wetware computing matter? #

• Biological learning. Neurons learn from few examples and adapt continuously, properties that AI research still struggles to reproduce efficiently.
• Energy. Brains are far more power-efficient than the hardware used for large AI models. Whether wetware systems can capture any of that efficiency is an open question discussed in biocomputing vs silicon.
• Science. Wetware experiments test theories of learning, such as the free energy principle, in human tissue.
• Access. Remote platforms, like the one described by Jordan et al. (2024), let computer scientists run experiments without their own cell culture labs.

Wetware computing vs neuromorphic computing #

Neuromorphic computing builds silicon chips that imitate how neurons spike and connect, and these chips often run spiking neural networks. Wetware computing uses actual neurons. Neuromorphic hardware is reproducible and mass-manufacturable; wetware offers genuine biological dynamics that silicon only models. The two fields inform each other, and both differ from deep learning on GPUs.

What are the limitations of wetware computing? #

• Drift and lifespan. Tissue matures and ages, so responses change over weeks.
• Variability. No two cultures are identical, which makes benchmarking hard.
• Bandwidth. Electrodes reach only part of the tissue, especially in 3D organoids.
• Life support. Cells need nutrients, gas exchange and sterility, often handled by incubators and microfluidics.
• Programming. There is no compiler; tasks must be translated into stimulation and feedback.
• Data standards. Results are easier to compare when shared in open formats such as Neurodata Without Borders.
• Ethics. Work with human neurons raises questions of donor consent and potential moral status, discussed in the ethics of organoid intelligence.

Where is wetware computing heading? #

The organoid intelligence roadmap by Smirnova et al. (2023) points to bigger, longer-lived tissue, 3D electrode interfaces, shared datasets and embedded ethics. Software infrastructure, including protocol tools, closed-loop runtimes and experiment records, is a large part of that work. It is the layer we focus on at Organoid Programming.

Frequently asked questions

What does wetware mean in computing?

In computing, hardware is the physical machine and software is the code. Wetware refers to living biological material, especially neurons, that performs computation. The word was used for decades in science fiction and to describe the human brain, and it is now used in research to describe systems that combine cultured neural tissue with electronics.

Is wetware computing the same as organoid intelligence?

They overlap strongly. Wetware computing is the broader term for computing with any living neural tissue, including flat 2D cultures of neurons. Organoid intelligence is a more specific research program focused on three-dimensional brain organoids grown from human stem cells. Most organoid intelligence work is wetware computing, but not all wetware computing uses organoids.

Can I run experiments on wetware remotely?

Some research platforms now allow it. Jordan et al. (2024) described a Neuroplatform that gives external researchers remote access to brain organoids on electrode arrays through software interfaces. Remote access lowers the barrier for computer scientists who do not have a wet lab, though experiments still depend on biologists who maintain the tissue.

How long do wetware computers last?

Living neural tissue does not last indefinitely. Cultures and organoids can be maintained for weeks to months with careful feeding, temperature control and sterile handling, but their activity changes as they mature and age. This drift is one of the main engineering challenges, because the same stimulation can produce different responses over time.

Sources

1. Jordan et al. (2024). Open and remotely accessible Neuroplatform for research in wetware computing. Frontiers in Artificial Intelligence.
2. Kagan et al. (2022). In vitro neurons learn and exhibit sentience when embodied in a simulated game-world. Neuron.
3. Cai et al. (2023). Brain organoid reservoir computing for artificial intelligence. Nature Electronics.
4. Smirnova et al. (2023). Organoid intelligence (OI): the new frontier in biocomputing and intelligence-in-a-dish. Frontiers in Science.
5. Thomas et al. (1972). A miniature microelectrode array to monitor the bioelectric activity of cultured cells. Experimental Cell Research.
6. DeMarse et al. (2001). The neurally controlled animat: biological brains acting with simulated bodies. Autonomous Robots.
7. Lancaster et al. (2013). Cerebral organoids model human brain development and microcephaly. Nature.