Traditional computational benchmarks of FLOPs or MIPS do not reflect the way the human mind “computes”. We believe a better way is to measure an architecture’s inherent capability to perform decisions or comparisons per second. We can easily relate intelligence to being a sequence of decisions and actions, based on changing inputs over time.
This simple description of intelligence is also a description of the Natural Intelligence NNP. The Natural Neural Processing Unit (NNP) receives input, determines if the input is relevant and, if relevant, decides what to do next. The NNP natively processes non-deterministic algorithms or pattern matching machines and processes these in parallel. In the Natural Neural Processor (NNP), these non-deterministic machines can be dissimilar and divergent threads that identify changing patterns in streaming data and all machines are operating independently on the data stream. The NNP allows a massive number of these non-deterministic machines to be operated in parallel, even though they are dissimilar in structure and in function. The processing threads these machines represent may freely diverge without compromising performance. These characteristics of the Natural Intelligence NNP implement a logical processing model analogous to the one embodied in the human neocortex.