DifferenceNeuronVsANN

Neurons vs. AI

Today's artificial intelligence may be inspired by the brain, but it isn't biologically grounded. It's like when a movie starts off saying inspired by true events and then goes off in whatever direction the writers and directors want. Today's artificial intelligence, particularly artificial neural networks and large language models are great, but bear virtually no relation to how our brains function.

Our brains operate on only 20 watts of power, less than my laptop, yet brains excel at context awareness, abstraction and common sense, areas where AI often fail.

The propagation delay, which is the speed that signals pass between neurons, is typically around a millisecond. And even under ideal conditions, a neuron can only repeat spikes slower than every 4 milliseconds because of its refractory period. Relative to electronics, these times are laughably slow. This has a huge impact on what sorts of AI processes are possible or impossible in neurons.

Any AI algorithm which relies on floating point numbers is out the window. At best, a given signal might take on 10 discrete values. More than that, a spiking rate signal is just too slow to be useful. Likewise for synapses, as I covered in a previous video, if your algorithm relies on floating point synapse weights, it is not neuraly plausible. Biological synapses are also limited to a few discrete values.

• Source: 2025-08-19 Neurons vs AI: They’re Nothing Alike