Organoid intelligence (OI): definition, how it works and why it matters

Organoid intelligence (OI) is an emerging research field that uses three-dimensional brain organoids, grown from human stem cells, as living computing hardware. Organoids sit on electrode arrays that deliver encoded electrical inputs and record the neurons’ responses, so researchers can test whether these networks learn, adapt and process information in ways that complement silicon computers.

Key takeaways

• Organoid intelligence treats living human neural tissue as a computing substrate, not just as a model of the brain.
• The field was framed as a formal research program by Smirnova et al. in Frontiers in Science (2023).
• An OI system combines four layers: the organoid, an electrode interface, encoding and decoding software, and a feedback loop.
• Early demonstrations, such as DishBrain playing Pong and Brainoware recognizing speech, are small proof-of-concept tasks.
• Major open problems include organoid size, lifespan, reproducibility, and ethics.

What does organoid intelligence mean? #

The phrase combines two ideas. An organoid is a small, self-organizing 3D tissue grown from stem cells that partly mimics the structure of an organ. A brain organoid is the version that develops neurons and supporting glia arranged in layers reminiscent of the developing cortex. “Intelligence” refers to the goal of the field: using that tissue to compute, learn and remember.

Smirnova and colleagues described OI in 2023 as “the new frontier in biocomputing and intelligence-in-a-dish”. Their paper treated it as a multidisciplinary program spanning stem cell biology, bioengineering, electrophysiology, machine learning and ethics. OI is a branch of the wider field of biocomputing, and it overlaps heavily with wetware computing, a term often used for the same work in industry.

For a long-form introduction, see our guide what is organoid intelligence.

How does organoid intelligence work? #

An OI experiment is a loop between living tissue and software. The building blocks are consistent across labs, even though the hardware differs.

1. Grow the tissue. Human induced pluripotent stem cells are guided into neural tissue over weeks to months. The method builds on the cerebral organoid protocol published by Lancaster et al. in Nature (2013).
2. Interface it. The organoid is placed on a microelectrode array (MEA), or on a high-density microelectrode array with thousands of recording sites.
3. Encode inputs. Information, such as a game state or an audio clip, is converted into patterns of electrical stimulation across selected electrodes.
4. Record outputs. The electrodes pick up extracellular voltage changes. Spike sorting and related analysis turn raw signals into spikes from individual neurons or groups.
5. Decode and feed back. Software interprets the activity as an output and, in closed-loop stimulation, sends new stimulation that depends on what the tissue just did.

Where does the learning happen? #

The hypothesis behind OI is that living networks change their own connections through synaptic plasticity, so their responses to the same input can shift with experience. Kagan et al. reported in Neuron (2022) that cultured neurons in a system called DishBrain improved at a simplified Pong game when given structured feedback, and they interpreted the result through the free energy principle. Our post on how neurons learn in a dish walks through that experiment in detail.

A second approach treats the organoid as a fixed but rich dynamical system. In reservoir computing, the tissue transforms inputs into high-dimensional activity, and only a simple readout is trained. Cai et al. reported in Nature Electronics (2023) that their “Brainoware” setup used this approach for speech recognition and nonlinear prediction tasks. Read more in reservoir computing with brain organoids.

Organoid intelligence vs artificial intelligence vs neuromorphic computing #

OI is often confused with AI and with neuromorphic computing. They share inspiration from the brain but differ in substrate.

The comparison shows why OI is pursued as a complement rather than a replacement. Living neurons may offer learning and efficiency properties that are hard to engineer, but they are slow, variable and difficult to program directly. The energy argument is covered in biocomputing vs silicon.

Why does organoid intelligence matter? #

Computing efficiency #

The human brain contains roughly 86 billion neurons, according to Azevedo et al. (2009), and runs on a small fraction of the power of a modern data center. Even modest steps toward capturing that efficiency would matter, because conventional AI is energy-hungry. OI is one way to study biological computation directly, in human tissue, rather than only imitating it in silicon.

Neuroscience and medicine #

The same platforms that test computation also reveal how human neural networks develop and fail. Organoids grown from patient cells let researchers compare learning and network activity between healthy and disease-linked genotypes, which could support drug discovery and the study of neurodevelopmental conditions.

Open access to wetware #

Access has been a bottleneck. Jordan et al. (2024) described a remotely accessible platform that lets external researchers run experiments on organoids over the internet. Remote, shared infrastructure is one route to more reproducible OI research.

What are the limitations of organoid intelligence? #

OI is promising but early. Honest limits include:

• Scale. Organoids are typically millimeters across and lack blood vessels, so inner cells can starve. Organoid vascularization and microfluidics are active research areas.
• Lifespan and drift. Living tissue changes over weeks and months, so a “program” may not behave the same tomorrow.
• Reproducibility. No two organoids are identical, which complicates benchmarking.
• Interface bandwidth. Flat electrode arrays touch only the surface of a 3D tissue.
• Programming model. There is no compiler for neurons. Tasks must be translated into stimulation patterns and rewards, which is a research problem in neural encoding and decoding.
• Ethics. Questions about consent from cell donors and the possibility of sentience must be addressed as the field grows.

What are the ethical questions around organoid intelligence? #

The Baltimore declaration, published by Hartung et al. in Frontiers in Science (2023), argued that OI research should be guided by “embedded ethics”, meaning ethicists and the public engaged alongside the science. Key issues include informed consent for donated cells, how to detect any morally relevant capacities in organoids, and how to share data responsibly. These topics are explored in the ethics of organoid intelligence and on our ethics page.

Where is organoid intelligence heading? #

The roadmap set out by Smirnova et al. points to larger and longer-lived organoids, 3D electrode interfaces, better analysis tools, and shared data standards such as Neurodata Without Borders. Progress will likely be incremental and measured in better benchmarks, not sudden breakthroughs. Our post on the future of biocomputing covers the open problems.

Building the software layer for organoid intelligence (protocols, closed-loop runtimes, and provenance tracking) is the problem we are working on at Organoid Programming.

Frequently asked questions

What is organoid intelligence in simple terms?

Organoid intelligence is the idea of using small clusters of living human brain tissue, grown in a lab from stem cells, as part of a computer. The tissue sits on a chip covered in tiny electrodes. Software sends electrical patterns in and reads the responses out, and researchers study whether the tissue can learn to process information in useful ways.

Who coined the term organoid intelligence?

The term was set out as a research program in a 2023 paper in Frontiers in Science by Smirnova and colleagues, led by researchers at Johns Hopkins University. The paper described organoid intelligence as a new frontier in biocomputing and proposed a roadmap covering organoid scale-up, electrode interfaces, machine learning for data analysis, and embedded ethics.

Is organoid intelligence the same as artificial intelligence?

No. Artificial intelligence usually runs as software on silicon chips such as GPUs. Organoid intelligence uses living neurons as the computing substrate itself. The two are linked, because machine learning is used to encode inputs and decode organoid responses, and some researchers hope living networks could eventually complement AI in learning efficiency and energy use.

Are brain organoids used for organoid intelligence conscious?

There is no evidence that current brain organoids are conscious. They are millimeter-scale tissues without sensory organs, blood supply or the large-scale structure of a brain. Researchers and ethicists nonetheless treat the question seriously, and the 2023 Baltimore declaration called for ethics to be built into the field from the start rather than added later.

What can organoid intelligence do today?

Today OI is a research field, not a commercial technology. Published work includes cultured neurons learning to play a simplified Pong game in closed loop, and a brain organoid used as a reservoir for a simple speech recognition task. These are proof-of-concept results on small tasks, and they do not yet compete with conventional computers.

Sources

1. Smirnova et al. (2023). Organoid intelligence (OI): the new frontier in biocomputing and intelligence-in-a-dish. Frontiers in Science.
2. Hartung et al. (2023). The Baltimore declaration toward the exploration of organoid intelligence. Frontiers in Science.
3. Kagan et al. (2022). In vitro neurons learn and exhibit sentience when embodied in a simulated game-world. Neuron.
4. Cai et al. (2023). Brain organoid reservoir computing for artificial intelligence. Nature Electronics.
5. Jordan et al. (2024). Open and remotely accessible Neuroplatform for research in wetware computing. Frontiers in Artificial Intelligence.
6. Lancaster et al. (2013). Cerebral organoids model human brain development and microcephaly. Nature.
7. Azevedo et al. (2009). Equal numbers of neuronal and nonneuronal cells make the human brain an isometrically scaled-up primate brain. Journal of Comparative Neurology.