Thanks BlueAnt :) I have to say (and I haven’t been saying much...

Thanks BlueAnt I have to say (and I haven’t been saying much for a while) but Anil to me is the most interesting listen of all the listens (once you get used to the way he speaks). He mention Tesla and 8 camera standard compute there a couple of times and how inefficient current cnn tech is at doing the inference job then he said a Neuromorphic based camera in a car could be trained for a specific task e.g to look for just street signs and only send data to the cpu if it sees a sign it has learnt. Think about that for a while. If cameras/sensors could be that fine-grained for specific tasks - potentially the car would not be limited to containing 8 always on video cameras, but could have hundreds of specifically trained (learned) sensors that would not fail at their allotted and simplified tasks and would only ever send meaningful data to be processed elsewhere. Therefore in this scenario a car could potentially contain hundreds of Akida chips instead of just dozens. .