A group of researchers attempted to connect brain tissue to computers in an experiment dubbed "Brainoware." At the heart of the system lay a brain organoid derived from human cells; however, instead of being transplanted into an animal, it was placed onto an electrode array. Through these electrodes, a computer could send electrical stimuli to the tissue and record how its neurons responded.
On one side was conventional electronics; on the other, a network of living human neurons. Yet, simply exchanging signals was not the primary goal; the researchers sought to discover something far more intriguing: could this tissue participate in computation? To test this, they used something unexpected—human voices. Recordings of people pronouncing Japanese sounds were converted into electrical stimulation patterns and sent to the organoid. Each input triggered a response in the neural network, which the computer recorded and used to attempt to identify the speaker.
Performance was initially limited, but during training, the organoid's responses and functional connectivity evolved, improving its performance in the speech recognition task. The researchers also used the system to predict a nonlinear dynamic equation. This does not imply that a tiny brain inside the machine was learning Japanese—there is no evidence the organoid understood the meaning of those sounds. Rather, the researchers were exploring something far more fundamental—and perhaps for that very reason, highly useful: plasticity.
Biological neural networks do not always respond in the same way; with repeated stimuli, their connections and activity patterns can change. Brainoware harnessed these dynamics as part of a system known as "Reservoir Computing." Instead of a conventional processor performing all the necessary transformations on its own, some of them occur within the responses of the living neural network itself. This creates a rather curious reversal: for decades, we tried to make machines imitate the brain, creating artificial neural networks, neuromorphic chips, and algorithms inspired by how neurons function.
Now, some researchers are taking the opposite path; rather than trying to build neurons using silicon, they are incorporating real neurons into the computing process. This line of research has even been given a name: Organoid Intelligence. It remains an experimental field—an organoid is incomparably simpler than a human brain; it lacks a body and our sensory systems, and there is no evidence that these systems possess a human mind or consciousness—yet Brainoware demonstrated an important concept.
A machine does not need to be built solely from electronic components to perform computations; part of it can be biological. This is precisely where the story takes another turn, because neurons are complex—they must remain alive and depend on extremely specific conditions. However, it may not be necessary to build something resembling a brain to harness biology for computation. There is a much simpler molecule that already performs another essential function within virtually all our cells—storing information in extraordinary quantities: DNA. And that represents yet another avenue of research and inquiry.
