Researchers from Tianjin University and Southern University of Science and Technology say their tiny lab-grown brain, or brain organoid, could learn tasks faster than current AI models. They say it uses much less energy than AI, which needs a lot of power and data to learn.
Swiss startup FinalSpark is offering biocomputers made from human brain cells, or organoids, for rent at $500 a month.
These biocomputers are up to 100,000 times more efficient than traditional silicon-based hardware and could accelerate AI development by mimicking human brain learning processes.
Australian startup Cortical Labs is developing a so-called DishBrain.
The system integrates brain cells grown in a dish with electronics to control simple tasks and play games like Pong. The company says the biocomputer system can work alongside traditional AI models for improved learning. Real brain cells learn tasks faster and consume less energy than current AI models. The brain cells in the DishBrain could be trained to control robotic arms or other devices.
At Indiana University, researchers are growing tiny, three-dimensional brain models from human stem cells. They call the technology Brainoware .They’ve demonstrated voice recognition capabilities. The Brainoware system was trained with 240 audio clips of different speakers. The audio was converted into electrical signals and sent to the brain organoid chip.
After just two days of training the system accurately distinguished between eight speakers 78% of the time from just a vowel sound. The brain models connect to computers via tiny electrodes. They use a gentle method called low-intensity focused ultrasound to stimulate the brain models, helping them grow into complex networks of brain cells.
By forming complex networks, these brain models might help robots learn and adapt better to new situations, like humans do.
SOURCE MATERIAL: AI + Human Brain Cells: Next Generation Intelligent Machines September 2024 https://youtu.be/AqM7mIUGfQ0?si=I52UF580umkIZdf2
