Although the current $500 per month subscription fee prices it significantly more expensive than Netflix, the Neuroplatform recently announced by FinalSpark promises to provide researchers with remote access to a supercomputer made out of 16 human forebrain organoids, reprogrammable via Python API.^[1] Neuroplatform this presents a significant leap in the access to advanced biocomputing, making it possible to comfortably link your laptop to a micro-brain that can run custom artificial neural networks.

The history of neuronal substrates in computer engineering dates back to 1999 when a team led by Emory-Georgia Tech professor Bill Ditto developed a simple computer capable of counting sums of numbers, made from neurons transplanted from leeches (hence called leech-ulator). Since then, the technology of making brain-like computational systems progressed towards the cultivation of complex systems made out of synthetic neurons grown from stem cells. Stem cells themselves represent a fascinating field of contemporary biomedical research, thanks to their pluripotency - they can be prompted to develop into any kind of cell in our body, e.g. skin cells, muscles, or organ tissues.

This feature of stem cells is used in bioengineering to build customized microphysiological systems, such as those built out of neurons. The neuronal networks grown from stem cells are called cerebral organoids: you can think of them as proto-brains or mini-brains that possess architecture and properties similar to those of mature brains. According to neurologist Thomas Hartung, the use of stem cells represents the most promising pathway for the development of neurocomputational substrates, making biological systems more computer-like by means of exploiting what he calls organoid intelligence (OI) or intelligence-in-a-dish - the capability of cerebral organoids to learn and solve problems.^[2]

The average weight of the human brain is about 1.5 kg, and its energy consumption is only 20 W (the equivalent of an LED light bulb). Despite the processing speed of neurons is extremely slow (only somewhere at 200 Hz, compared to several GHz of clock speed delivered by commercial CPUs), the brain’s advantage is in massive process parallelism, which allows the compound computing power of the brain to match the world’s fastest supercomputers (which operate at around 1 ExaFLOPS). Moreover, compared to the energy consumption of these supercomputers (such as 20 MW for the supercomputer at Oak Ridge National Laboratory), human brains are extremely energy efficient.

For this reason, organoid computers are an attractive alternative to standard silicon-based computers, especially given the fact that the computational and data storage demands of computers skyrocketed since the mass adoption of large language models and other forms of advanced AI. Since the greening of the global economy requires not just a transition to renewable energy resources, but also a significant increase in energy efficiency, biocomputing with brain organoids can help to keep the world’s information economy running at much lower energy costs.

Looking closer at organoid intelligence, even simple 2D neuronal circuits composed just of a few cells can learn to play games and run spatial simulations. Thanks to the natural features of neurons, these 2D circuits have the potential not just to store or process information, but also to improve their operation, which makes them trainable similarly to state-of-the-art artificial neural networks. Essentially, such organoid ensembles are self-improving, intelligent systems. More advanced brain organoids are structured in 3D networks with fifty to hundred thousand individual neurons and only several millimetres in size. While these are still relatively tiny, the upcoming generation of organoids may grow up to the size of 1 cm, such as brain organoids developed at the Johns Hopkins University SURPASS project. The information-processing capacities of these organoids make them great adepts for becoming the future computational substrate of AI, offering at the same time great time and energy savings.^[3]