Artificial neural networks present many similarities with the mechanisms underlying the functioning of the human brain. As we have already highlighted, the evolution of future artificial intelligence could pass through the use of cerebral organoids, three-dimensional aggregates of cells grown in the laboratory that try to recreate the structure and function of the brain in a simplified way. From here to the birth of a computer bioibridowhich integrates biological components with electronic ones, the step could be short.
This is demonstrated by a team of academic researchers led by Feng Guo (Indiana University, Bloomington), who succeeded in creating a configuration based on the use of a brain organoid made up of stamina cells human and connected to an electronic chip, in order to perform and record simple computational tasks.
The project is called Brainoware and clearly shows the boundless horizons towards which bioinformatics could lead us in the future, i.e. that scientific discipline that combines more traditional information technology with interaction through the use of biological structures. The study has just appeared in the scientific journal Nature Electronics.
What is Brainoware and how does it work
In recent years, scientists and engineers have tried to get closer to the brain’s capabilities by designing hardware and algorithms that can approximate its structure and how it works. Known as computing neuromorficothis particular branch is making continuous progress but its processing is energy-intensive and the training of artificial neural networks requires time (in addition to a series of particular attention).
Guo and his colleagues took a different approach using vero brain tissue human grown in the laboratory. Stem cells were encouraged to develop into different types of brain cells that organized themselves into three-dimensional organoids, complete with interconnections and structures.
Brainoware consists of brain organoids connected to a microelectrode array high density: they constitute the interface between the brain tissue and the electronic system. The connectivity thus achieved allows the recording of neural activities and the transmission of electrical signals to the brain tissue.
The presented solution is based on a mechanism known as reservoir computing: cerebral organoids act as “reservoirs” (reservoir) in which information is processed, transported through electrical stimulations.
The layers of input e output they are implemented using ordinary computer hardware. The input layers receive data from the external environment or from the organoids, while the output layers interpret the neural responses processed downstream of the system reservoir computing.
By training the input and output layers to work symbiotically with the organoids, it is possible to induce the system to learn specific taskssuch as speech recognition or complex pattern prediction.
What the biohybrid computer created from brain tissue grown in the laboratory is already capable of doing
In the specific case of Brainowareresearchers have been successful in training the system to perform tasks such as speech recognition and the prediction of nonlinear models. For example, the biohybrid computer has succeeded in identify entries specifications with significant accuracy and to predict complex patterns such as those representable with a Hénon map.
The research team highlights how Brainoware can effectively process information and perform computational tasks without human supervision.
Among the various tests conducted, the system based on the use of brain organoids was recognized with one precision equal to approximately 78%, with just two days of training and using 240 pre-recorded audio clips, the voices of various people. Guo specifies that it is a first approximation of what will be possible to achieve in the future: “the results of this experiment constitute the first demonstration of the use of brain organoids in the IT field. We were able to confirm the potential of brain organoids in the future of bioinformatics“.
In parallel, however, Lena Smirnova and Brian Kafo (Johns Hopkins University) talk about ethical issues which will arise as experts develop increasingly sophisticated uses of brain organoids. However, as well as on the artificial intelligence side, “the research provides fundamental insights into the mechanisms of learning, neurodevelopment and cognitive impact of neurodegenerative diseases. It has great potential to produce new results that could be useful for developing models to test new treatments“.
Opening image credit: iStock.com/imaginima