Small pieces of organs have been used in research for some time in combination with chips to directly obtain information about their functioning under the influence of, for example, medicines. Research into brain tissue is also regularly conducted. We have now even managed to get a bunch of brain cells to do some calculations.
At Indiana University Bloomington in the US, a team led by researcher Feng Guo has managed to grow single brain cells from stem cells and then link them to a computer chip. This has been successful before, but this team then linked the brain-computer interface to an AI tool. It subsequently turned out to be possible to perform some calculations with this system, but also to learn and remember things, and even do a little bit of speech recognition. The research was published in Nature on December 11.
The Indiana University Bloomington campus.
It is a first major step towards the use of brain tissue for certain types of calculations. While standard computers are very good at calculations, biological computers are a lot better, and especially more efficient, at processing more complex data sets. A human brain only uses between 100 and 150 watts while constantly processing large visual and audio data streams. An efficiency that traditional computers cannot come close to when it comes to processing large amounts of data.
The goal of Brainoware, as the researchers call their brain-computer interface, was to use the brain cells to send and receive data. This worked, but the researcher also found changes in the neural pathways in the tissue, which would mean that the data was processed in a certain way. The team then tried to give the piece of brain tissue tasks, such as solving some mathematical calculations. Then they tried a speech recognition benchmark test, where 240 audio clips were played of eight different people pronouncing Japanese vowels. This first had to be converted into electronic signals and when the brain had finished it had to be decoded by the AI tool. It was clear that the cells could decode the audio fragments, and improved after some training. However, the brain-AI combination did not exceed an accuracy of 78%, which is not as good as artificial neural networks. However, further steps are still far away, because it is difficult to keep brain organoids alive and make them larger in a lab.
Source: Technology Review