Competitive landscape, page-46

@shareman yes you are correct to a certain extent, where if cost is an issue there will be other chips used in applications that do not require a dedicated SNN. I do know the competitive landscape as I actually did a somewhat in-depth comparison before putting my money in BRN. That being said, the future for edge devices will be in SNNs and other neuromorphic technologies that closely mimic the neurons in a human brain.

Computational models inspired by the brain’s ability to naturally encode and process information in the time domain. This is modelled in Spike-timing-dependent plasticity (STDP) which is a biological process that adjusts the strength of connections between neurons in the brain. The process adjusts the connection strengths based on the relative timing of a particular neuron's output and input action potentials (or spikes), meaning that something which occurs more often is "registered" in the brain strongly, and also that if there is a sudden change in a static environment (like in vision applications when someone walks into the scene of a room), the "brain" model would cascade spikes. The added temporal dimension is believed to render them more computationally efficient than the conventional artificial neural networks. Brainchip owns intellectual property (IP) in STDP on silicon https://brainchipinc.com/brainchip-bicc2020-presentation/.

Because of the way processors with von Neumann design work, competing von Neumann-based processors would have to analyse every frame -- often on the cloud, which consumes additional power in transmission of the video feed as well as on the cloud server side -- even though the scene does not change, which means it continuously consumes power and is processing regardless of what is in frame. A SNN-based processor like Brainchip's Akida would ignore the static scene and only "spike" only when the scene changes, and thus "wake" the rest of the circuitry to record or send an alert or whatever it is programmed to do. Hence, there is no commercially available SNN chip that OEMs can use with extremely low power consumption (mW to µW)... except for Brainchip's Akida -- a fact attested to by Mike Davies the head of Intel's Loihi project I believe -- IIRC he said something to the effect of Loihi is an experimental chip, if you want a commercial chip go to Brainchip (I can't find the related post so will have to rely on someone else to link it).