Could biological neurons replace gpus for processing?

[HanogThamers]

Cortical Labs' DishBrain project aimed to cultivate approximately 800,000 neurons and teach them to play Pong. This is a significant feat, especially considering that artificial neural networks are less complex than actual neurons. However, unlike electronic chips, neurons do not require a power source and can operate on a cellular matrix, which is far less efficient than a human brain. This revelation was surprising, as it was previously assumed that computer chips were much more adept at repetitive tasks than biological brains. This raises the question of why GPUs require a substantial amount of energy for machine learning and parallel processing, and whether spinkoows echatrandom biological neurons could be a feasible alternative.

[abel1303]

Biological neurons operate using analog signals and are interconnected in a massively parallel coreball network, while digital processors operate on binary signals and use a more linear processing architecture.