Competitive landscape, page-2348

It's good to see that edge AI technology is evolving at a rapid pace. Currently everything seems better than the usual CPU platform and it's good to see that devices and systems are adopting AI functionality. But as the edge AI space expand, there will be a need a platform that can have cross compatibility, high scalability, multi-modality, high speed of processing data and learning and low energy consumption.
Brainchip's Akida SNN platform appears to tick all the boxes.
Only pure mesh digital SNN can have a compact stand alone chip with the potential neuronal and synaptic capacity to match that of the brain. Any other platforms will need stacking which will always compromise on size and efficiency...and just doesn't pass the "future" test! Currently only Akida's SNN can have the structure to have its 1.2M neurons and 10B synapses capacity...that's plenty for current needs and plenty of gap from its competitors i.e, Loihi and TrueNorth. In fact if you look into the architecture of the cores in Loihi and TrueNorth, they simply will never get to Akida's current neuronal and synaptic capacity unless they start stacking or increase in size! Sooner or later, the amount and complexity of data needing to be processed at the edge will demand for more neuronal and synaptic capacity just like RAM for CPU...and you cannot just keep stacking chips! As the chip foundries move into sub 4nm architectures, Akida's neuronal cores or nodes will also shrink which will move Akida closer and closer to the capacity of the human brain.
Only Akida's platform appears to be able to be programmed to be cross compatible with any other platform. When everyone is happy playing around with various ANN/CNN platforms and looking into the "next generation" SNN platform especially if other competitors begin to use SNN, who will they turn to?...only Akida...Loihi and TrueNorth just does not have this capacity to allow a smooth transition without completely new programming and algorithm. Compatibility issues will be a big problem if there are many platforms. This also is related to the different modality of use. It will be ridiculous for a car or a robot to have several different edge AI platforms for different AI sensory modalities and unable to communicate effectively with each other. Currently only Akida can run every sensory modalities at once plus extended to run other sensors such as radiation. Of course now with only the beginning of AI, devices are generally mono-modal eg, ear plugs can only listen...but just imagine ear plugs can also have side vision to warn you of in coming traffic or assailant when you are jogging! Multi-modality will definitely be the future in the edge AI space!
The speed of data processing and learning at the edge basically will depend on real-time, event based and instantaneous processing and learning. How fast this occurs depends on the amount of synaptic capacity which determines the instantaneous in-memory required for instantaneous processing and learning. Akida can process data inputs instantaneously and learn with "one shot" (or pass)...apparently other platforms can only do things sequentially and with multiple passes.
Being pure SNN, Akida will always have the lowest energy consumption...no disputing this fact.
So all it is now is that the proof is in the pudding and as people play around with edge AI and the space becomes much more complex and multi-modal, only Brainchip's Akida platform can handle this test of time IMO. DYOR.. you will surely come up with the same conclusion as I have