Scientists have developed a system that combines brain-like tissue with electronic hardware for speech recognition and calculation. This system advances research on powerful biological computers.
This work is a boost to efforts to develop so-called Neuromorphic Computing Devices powered with human brain cells, which have the potential to be faster at learning and more energy efficient than traditional silicon-based computers. Researchers say that a breakthrough in the emerging field of artificial intelligence could enable advancements in medical science and treatment.
The authors of an article on the hybrid creation, published on Monday in Nature Electronics, wrote that “Brain-inspired computer hardware aims at emulating the structure and operating principles of the human brain and could address current limitations in AI technology.”
Researchers at Indiana University Bloomington and the Universities of Florida and Cincinnati as well as the Cincinnati Children’s Hospital Medical Center developed the “Brainoware”, a system that uses a brain organoid, a three-dimensional neural structure made from human stem cells.
The team used computer hardware to send electrical stimuli to the organoid, and then read the neural activity produced by the organismoid in response. The system has the ability to identify Japanese vowel sounds and accurately anticipate a mathematical correlation.
The scientists asked Brainoware to distinguish between eight male Japanese speakers on a total 240 audio clips. After training its algorithm, the system improved from approximately 51 per cent to approximately 78 percent accuracy.
Researchers tested the system in a maths challenge. They asked it to predict the Henon map – a chaotic representation. Brainoware was slightly less accurate than silicon-based networks in the test, but it had a training time 90% lower.
This research follows the February launch by an international team led by scientists at Johns Hopkins University, US, of , a detailed roadmap towards “organoid Intelligence”. The plan promises breakthroughs in computing, neuroscience, and other areas of research.
Brainoware’s experiments show the potential of the field, even though researchers acknowledge that a general-use biological computer may still be decades off.
In a comment published in Nature Electronics on Monday, a group of Johns Hopkins researchers wrote that biological computing research will likely yield “foundational insight into the mechanisms for learning, neural development, and the cognitive implications associated with neurodegenerative disorders”.
The three researchers who did not participate in the Brainoware study said that “in the next few decades, it is likely that increasingly complex neural systems will emerge, which can interact with increasingly complicated artificial environments.”
Creating brain-like intelligence raises ethical questions about generating consciousness in a lab.
The researchers said that as the complexity of these organoid system increases, the community must examine the neuroethical issues surrounding biocomputing systems that incorporate human neural tissue.