According to a recent New Scientist article, engineers at Indiana University Bloomington have developed a biocomputing system named "Brainoware" that consists of living brain cells linked to a computer. The brain organoids, which are small lumps of nerve cells, were trained to perform basic speech recognition by identifying the voice of one individual from a set of 240 audio clips. The system demonstrated an adaptive learning process, improving accuracy from 30-40% to 70-80% after two days of training without feedback.
The researchers aim to explore the potential of biocomputing using living nerve cells as an alternative to conventional AI systems with high energy consumption and inherent limitations of silicon chips. However, challenges include the simplicity of the task and the limited lifespan of the organoids, which can only be maintained for one or two months.