Brainoware: This happened when scientists combined real human brain tissue with electronics

The brain is amazing because neurons do two jobs: processing and remembering. Regular computers keep these jobs separate, but Brainoware is different. It combines real human brain tissue with electronics to make a new kind of computer.

Behind Brainoware is a team at Indiana University Bloomington, led by Feng Guo. They tested it with tasks like recognising speech and predicting equations. Brainoware wasn't as accurate as a computer with just hardware and artificial intelligence, but it's a big step toward a new kind of computer.

Ethical issues

Add WION as a Preferred Source

Even though Guo and his team followed ethics rules, other researchers from Johns Hopkins University say we need to be careful. Lena Smirnova, Brian Caffo, and Erik C. Johnson, who didn't work on this study, say we must think about ethical issues when using human brain tissue in computers.

Before Brainoware, the best attempt to copy the brain was in 2013 with Riken's K Computer. It was one of the most powerful computers globally but took 40 minutes to simulate one second of just 1-2 per cent of the brain.

How was the brain tissue obtained?

Scientists have tried to make computers like the brain using neuromorphic computing, but it uses a lot of energy and takes a long time to train. Guo and his team tried a different way. They used real human brain tissue grown in a lab, forming mini-brains called organoids.

These organoids aren't real brains – they're just tissue structures without thoughts or feelings. They help study how the brain works without using a real human brain. Brainoware links these organoids to microelectrodes using a type of artificial neural network called reservoir computing.

Watch:How alcohol impacts the brain | Brain can repair damage in 7.3 months

Electric signals send information into the organoid, where it's processed before Brainoware gives its answers as neural activity. Regular computer parts handle the input and output. These parts had to be trained to work with the organoid.

Results

To test Brainoware, the researchers gave it 240 audio clips with Japanese vowel sounds from eight speakers. They asked Brainoware to identify one speaker's voice.

After just two days of training, Brainoware could identify the voice with 78 per cent accuracy. It was a bit less accurate than other artificial networks, but it reached similar results in much less training time.

(With inputs from agencies)