Scientists have made a groundbreaking achievement by merging brain-like tissue with electronics, creating an organoid neural network capable of voice recognition and solving complex mathematical problems. This development extends the field of neuromorphic computing, which involves modeling computers after the human brain, by incorporating actual brain tissue into a computer system.
The study was conducted by researchers from Indiana University, the University of Cincinnati, Cincinnati Children’s Hospital Medical Centre, and the University of Florida. Their findings were published on December 11, 2023, marking a significant advancement in multiple areas of science and engineering.
The rise of artificial intelligence (AI) has led to the development of artificial neural networks, which are brain-like networks of neurons created with silicon chips. However, these networks face a challenge because the memory units and data processing units are separate. This results in increased time and energy demands as the system continuously moves between these units while operating.
To address this issue, scientists have been exploring the use of biological neural networks, which are networks of live brain cells. Unlike AI hardware, brain cells store memory and process data without physically separating the two. The brain only requires 20 W of energy to perform the same amount of work that AI hardware would need 8 MW for. This significant difference is due to the seamless integration of memory and data processing in brain cells.
In the field of biocomputing, researchers are utilizing biological components to perform computational processes. In a previous study, Australian researchers cultured brain cells and trained them to play a table-tennis-like video game, demonstrating the early steps of long-term training. In the recent study, the US researchers used brain organoids, which are aggregates of brain cells, to create an organoid neural network capable of recognizing speech and solving complex mathematical problems.
Brain organoids are created by extracting human pluripotent stem cells and developing them into brain cells. These organoids consist of different types of brain cells, including neuron progenitor cells, early-stage neurons, mature neurons, and astrocytes. The researchers connected the brain organoid to an array of microelectrodes, forming an organoid neural network. They then incorporated this network into a system called a reservoir computer, which contains input, reservoir, and output layers.
The input signals are sent to the reservoir, which converts them into mathematical entities for processing. The output is a readout from the reservoir. The researchers named this system Brainoware. They demonstrated its capabilities by predicting a mathematical function known as a Henon map and distinguishing between different Japanese vowels based on audio clips. Brainoware achieved remarkable accuracy with less training compared to artificial neural networks.
While Brainoware still has limitations and ethical considerations, it represents an innovative step towards the development of biocomputing systems. Future research could focus on optimizing input encoding, maintaining uniformity of organoids during longer experiments, and tackling more complex computing problems. The study also provides valuable insights into learning mechanisms, neural development, and the cognitive implications of neurodegenerative diseases.
In conclusion, the fusion of brain-like tissue with electronics to create an organoid neural network has opened up new possibilities in tissue engineering, electrophysiology, and neural computation. This breakthrough in biocomputing has the potential to revolutionize the field of artificial intelligence and pave the way for unprecedented advancements in technology.
