Ideal AI will use Human Brain Cells

Researchers at Johns Hopkins University and Cortical Labs in Melbourne are trying to make a new kind of computer that uses human brain cells. According to Cortical Labs, this computer could be better than electronic computers for some tasks while using a small fraction of the electricity that computers and server farms use today. In a recent article in Frontiers in Science, the team describes their plan for "organoid intelligence" (OI), in which stem cells would be used to grow clusters of 50,000 brain cells called "organoids." They want to get to about 10 million neurons, which is about how many there are in the brain of a tortoise.

The researchers think that biological computers could do certain tasks better than electronic computers. For example, they could learn to tell the difference between two kinds of objects from just a few examples, which would take many thousands of AI algorithms to do with electronic computers. Also, biological computers use less energy than electronic ones. For example, our brains can store the same amount of information as a million average home computers but only use a few watts of power.

The team at Cortical Labs has already shown that human brain cells can be used to make biocomputers. For example, in a recent paper published in Neuron, brain cells in a flat culture learned to play Pong. Researchers think that this technology could also help them study personalized brain organoids made from skin or small blood samples from people with neurological diseases like Alzheimer's and run tests to find out how genes, medicines, and toxins affect these diseases.

Researchers think that the new field of biocomputing will bring about huge improvements in speed, processing power, data efficiency, and storage capacity, all while using less energy. They think that the spirit of openness and cooperation that led to their collaboration is important for success and is a key touchpoint for industry collaboration.