Brainchip VS IBM Synapse, page-7

It is hard to pin point the exact difference without expert technical knowledge in the field. If you search the web you will find there are quite a few experts bagging TrueNorth, one neuroscientist was very vocal. Peter Van Der Made himself would know more than anyone else in the field. In comparing who's got more people and money lets not forget that Peter was the Chief Technology Officer at IBM when he decided to leave and develop Brainchip. He was appointed CTO after IBM bought a computing immune system that was invented by him.

My understanding of TrueNorth so far is that it still falls under the 'neural software computing' category. It requires massive computing resources to get it to work in a limited way. It has frustrated even their own team's programmers with its own set of programming languages and codes. It also seems they have over simplified the mimicking of the human brain's neuron and synaptic functions - to the point where each synapse function is represented only by a point on the chip's program (something like that) whereas in real life the function of a synapse is a lot more complex than that. True North is also extremely expensive and their applications will only be useful in limited areas. 1 million neurons and 250 million synapses probably means little if their fundamental architecture is flawed. Peter have said that if you are trying to emulate the biological brain, the complexity to emulate it reaches infinity the harder you try if the fundamental architectural difference is flawed. Meaning that if you want to emulate the brain's functions then you really have to understand the workings of the brain down to the synapse level and copy it from there - what he calls reverse engineering the brain.