It's funny that none of the AI firms in the limelight are making any attempt to actually copy the architecture of biological neutral networks. Analog and neuromophic neural network hardware acceleration chips, ie hardware that mimics how biological brains compute, have both crazy energy efficiency and high performance. The only two firms that have actually been making a serious effort at developing such hardware are Intel and IBM. For both companies neuromophic chips have been r&d projects that have already had over a decade of resources poured into and will probably require at least another decade before producing something commercially viable.

It's also evidence that firms like OpenAI, Google, Microsoft, and all the AI startups that are spouting off about working on creating actual AI are full of shit. None of them are working on developing hardware to run real neural networks able to emulate the type of intelligence human brains are capable of, ie dynamic, general, and realtime learning.