Indiana University's 'Brainoware' uses human brain cells to slash AI energy use

A team at Indiana University Bloomington, led by Feng Guo, has developed 'Brainoware' – a biological computing system that uses tiny clusters of lab-grown human brain cells (brain organoids) to perform mathematical computations. These 3D neural networks, each about the size of a lentil and lacking blood vessels, successfully solved a complex nonlinear equation known as the Hénon map. The work, published in early 2023, demonstrates that living neural tissue can be trained to perform specific computational tasks, representing a tangible step toward biocomputers that blend biology with silicon.

The key advantage of 'Brainoware' is energy efficiency: a human brain operates on around 20 watts, while the AI that defeated Go champion Lee Sedol consumed over 5,000 watts. As AI systems grow more powerful – powering self-driving cars, autonomous aircraft, and increasingly complex decision-making – their energy demands threaten to become unsustainable. Brett Kagan from Cortical Labs, who previously worked on 'Dishbrain' using mouse brain cells, called organoids 'an exciting next step' but noted the published research needs more detail. Ethical concerns have also been raised about using living human brain tissue as hardware, especially as organoids become more sophisticated. Despite these questions, 'Brainoware' points to a future where AI might run on biological processors, dramatically cutting energy costs and carbon footprints.