The reason I bring it up is that the HoloLens is pretty limited. Given what I've seen it do with its limited power and the kind of computing power they're putting in Teslas, not to mention the extra sensors (RADAR), it seems more than adequate to perform real-time image processing, even in the short term. That's not even considering what's coming down the pipeline, like fleet learning, the new Nvidia GPUs, etc.
Also when the HoloLens throws a fit it's maybe a 5% margin of error in terms of how the image moves. You'd think an autonomous vehicle wouldn't crash because of that much inaccuracy considering human drivers don't normally with less accuracy than that.
