80,000 rat brain cells create a living computer

A computer based on neurons. Image source: New Scientist magazine website

According to a report on the website of the British magazine New Scientist on the 16th, American scientists used 80,000 living cells from mice to build a living computer that can simply recognize light and electrical patterns. This machine can be integrated into robots that also use living muscle tissue. The research team introduced this research at the March meeting of the American Physical Society.

In the latest study, the research team at the University of Illinois at Urbana-Champaign first cultivated approximately 80,000 neurons from programmed mouse stem cells in a culture dish, and then placed the neurons under optical fibers and on an electrode grid to allow them to receive electrical and light stimulation. All components were placed in a palm-sized box, which was placed in an incubator to keep the cells alive.

To train the neural computer to distinguish different signal patterns, the research team created 10 different patterns of electrical pulses and flashes, and played these patterns repeatedly for an hour, while using traditional computer chips to record and process the electrical signals generated by neurons. The results showed that neurons generated the same signal every time the same pattern appeared. In addition, the researchers also used reservoir computing to allow neurons and chips to work together to minimize the time and energy spent on identifying and processing signals.

To evaluate the device’s performance, the research team calculated a performance score called F1, which is often used to indicate how efficiently a neural network recognizes patterns, with 0 being the worst and 1 being the best. The device’s best score was 0.98.

The newly developed device can be integrated into robots made from living muscle tissue. Incorporating neurons into robots means that the neurons can sense the environment and then process these inputs all at once.

The researchers say that using living cells for computing, especially reservoir computing, could help create energy-efficient devices that can continue to work even if some of their components fail. Robots that combine living neurons and reservoir computing could therefore have advantages over traditional robots.