"However, it’s been shown that 'a single neuron may be able to compute truly complex functions.'"

Yikes! I hadn't yet seen this research. I suspect this punches a very big hole in the possibility of human-level intelligence any time in the near future, as it represents far more than a simple exponential acceleration of the current path of ANNs.

Am I wrong? I mean, unless it's some kind of optimization, whereby single dendrites learn to compress, or generalize, the activity learned by layers of neurons (in order to release those layers for other tasks), then we're a very long way from understanding how biological (human) brains actually work, no?