In a warm laboratory chamber, clusters of human cells can organize into miniature neural structures that fire electrical signals. These brain organoids are neither people nor complete brains, but their activity is becoming sophisticated enough to support disease research, drug testing and a more speculative idea: using living neurons as a computing substrate.
The two ambitions share the same biological foundation but demand different standards of evidence. Organoids can model aspects of human neural development more directly than many animal experiments. Claims that they represent a practical alternative to silicon—or possess anything resembling consciousness—remain far less settled.
Living models offer a new view of brain disease
Researchers can create organoids by reprogramming cells from adult skin, blood, hair or teeth into induced pluripotent stem cells and guiding them toward neural development. The process avoids taking tissue directly from a brain and allows scientists to build models linked to a particular donor’s genetics.
At the University of California San Diego, Alysson Muotri’s laboratory grows organoids containing around five million cells, including roughly 2.5 million neurons—comparable in cell count to a bee brain. The team has used them to investigate autism, including cultures derived from Muotri’s son.
Over eight months at body temperature, some organoids produced increasingly repetitive electrical oscillations resembling patterns observed in premature babies. Similarity in a signal does not make an organoid equivalent to a developing child. It does show that organized neural activity can emerge without the anatomy or sensory experience of a complete brain.
This is where the near-term medical value is strongest. Researchers describe failure rates approaching 95 percent for neuropsychiatric drug trials, a problem partly attributed to models that do not reproduce human biology well. Organoids could provide a more relevant test bed for studying disease mechanisms, toxicity and how compounds affect learning-related activity before clinical trials.
Neural cultures are entering computing experiments
Other groups are treating living neurons as adaptive information processors. Johns Hopkins researchers are exploring biocomputing and training cultures on game-like tasks to study how drugs influence learning and memory. Cortical Labs has demonstrated neural cultures interacting with Pong and Doom and now offers a system called CL-1.
The company’s devices keep cultures of up to one million neurons alive on electrode arrays for as long as six months. Electrical stimulation supplies input and measured neural activity becomes output. Cortical Labs markets remote access as “neurons as a service” and argues that living networks could become an energy-efficient, adaptable substrate.
These are company claims built on early experiments, not evidence that biological computers are ready to replace conventional AI hardware. Training, reliability, reproducibility and long-term maintenance remain formidable problems. A disputed description of a 2022 Pong-playing culture as sentient also illustrates how quickly technical demonstrations can outrun scientific agreement.
For now, the ability to change behavior in response to feedback is more useful as an experimental model than as proof of general intelligence. The cultures may help researchers observe learning in living tissue even if they never become commercially competitive computers.
Growth brings practical limits and ethical uncertainty
Most brain organoids remain only a few millimeters wide because they lack blood vessels capable of delivering oxygen and nutrients to a larger structure. Researchers are developing vascular and perfusion systems, with some aiming for organoids around one centimeter across, approximately mouse-brain size.
Removing that constraint could make models more useful, but it also sharpens ethical questions. Current organoids are generally treated as neither animals nor people and can fall outside federal animal-welfare rules. There is no consensus that they are conscious or sentient. There is equally no simple test that settles how growing size, connectivity and sensory input might change their moral status.
Policy and funding are already moving. In 2025, the US National Institutes of Health said it would no longer issue grants devoted exclusively to animal testing, and an $87 million center is intended to standardize organoid development.
That investment points toward the field’s most defensible promise: better and more consistent human tissue models. Experimental computing may produce valuable tools along the way, but living neural systems cannot be evaluated only by speed or energy use. Their scientific usefulness, biological fragility and uncertain ethical boundaries must advance together.
