Brainoware with unsupervised learning for AI computing. ASchematic of an adaptive reservoir computing framework using Brainoware. bSchematic diagram of the Brainoware setup that mounts a single brain organoid on a high-density MEA to receive inputs and send outputs. vsComprehensive immunostaining of cortical organoids showing complex three-dimensional neuronal networks with diverse brain cell identities (e.g., mature neuron, MAP2; astrocytes GFAP; neurons at early stage of differentiation, TuJ1; neural progenitor cells, SOX2). d,Scheme demonstrating Brainoware’s unsupervised learning hypothesis by reshaping the BNN during training, and the inhibition of unsupervised learning after blocking synaptic plasticity. Scale bar, 100 μm. Credit: Natural electronics (2023). DOI: 10.1038/s41928-023-01069-w
Groups of brain cells raised in a laboratory and connected to a computer are able to solve basic speech recognition and math problems.
Feng Guo, a bioengineer in the Department of Intelligent Systems Engineering at Indiana University in Bloomington, said his study is a major step in demonstrating how brain-inspired computer neural networks can do advance the capabilities of artificial intelligence.
Guo and his team developed bundles of specialized stem cells that developed into neurons, the main component of the brain. A typical brain is made up of 86 billion neurons, with each neuron connected to up to 10,000 other neurons.
The ball of neurons, known as an organoid, created in Guo’s lab is less than a nanometer wide. It was connected by an array of electrodes to a circuit board, where machine learning algorithms decoded the organoid’s responses.
The researchers named their creation Brainoware.
After a brief training period, Brainoware was able to distinguish the voices of eight subjects based on varying pronunciation of vowels. The system achieved an accuracy rate of 78%.
Brainoware was also able to predict a Hénon map, a mathematical construct in the field of chaotic dynamics, with greater accuracy than an artificial network.
“This is a first demonstration of the use of brain organoids (for computing),” says Guo. “It’s exciting to see the possibilities of organoids for bioinformatics in the future.”
One of the main advantages of bioinformatics is its energy efficiency. Currently, artificial neural networks consume several million watts of energy per day. The human brain, on the other hand, only needs about 20 watts to function for a day.
Brainoware is “a bridge between AI and organoids,” Guo said. “Organoids are like “mini-brains”.
“We wanted to ask the question of whether we can leverage the biological neural network within the brain organoid for computing. This is simply a proof of concept to show that we can do the work,” said Guo said.
A future application of bioinformatics systems is to study neurological diseases such as Alzheimer’s disease. The ability to harness cellular activity also opens the door to decoding brainwave activity during sleep and potentially recording dreams.
Challenges remain. Among them will be the task of keeping the organoids healthy and well-nourished, a 24/7 task.
And there are other concerns too.
“As the sophistication of these organoid systems increases, it is critical that the community examines the myriad neuroethical issues that surround bioinformatics systems incorporating human neural tissue,” Guo said.
“It may be decades before general bioinformatics systems can be created, but this research has the potential to generate fundamental insights into learning mechanisms, neuronal development, and the cognitive implications of neurodegenerative diseases.”
“We have a long way to go,” he added.
The study was published in Natural electronics.
More information:
Hongwei Cai et al, Calculation of brain organoid reservoir for artificial intelligence, Natural electronics (2023). DOI: 10.1038/s41928-023-01069-w
Lena Smirnova et al, Reservoir computing with brain organoids, Natural electronics (2023). DOI: 10.1038/s41928-023-01096-7
© 2023 Science X Network
Quote: Brain tissue on a chip enables voice recognition (December 12, 2023) retrieved December 12, 2023 from
This document is subject to copyright. Apart from fair use for private study or research purposes, no part may be reproduced without written permission. The content is provided for information only.
