Hi
@TechGirl
Perceive have a number of patents this one seems the most likely to provide the answers if
@Diogenese is so inclined:
Some embodiments provide a method for compiling a neural network program for a neural network inference circuit. The method receives a neural network definition including multiple weight values arranged as multiple filters. For each filter, each of the weight values is one of a set of weight...
patents.justia.com
My technophobe read tells me they are not using SNN, they train a network in advance and try to bring it to 4 bit activations for a single processing purpose and store it in memory on the and then use a CPU to send it to the network to process the input. There is a great deal more going on and it seems very complex and rigid in that whereas AKIDA can process all the senses plus radar, Lidar and ultrasonics and learn on the fly incrementally and through one shot learning you get the one trained function.
Peter van der Made spoke about how this low powered single purpose processing was possible back in 2019 at the AGM when discussing ETA’s chip.
I have the very lay impression their chip is more expensive to produce in the first instance and from the Careers on offer they seem to be targeting 7nm to 16nm which are definitely more expensive than AKIDA at 28 nm regardless of complexity.
My opinion only DYOR
FF
AKIDA BALLISTA