Except EyeQ4 is still planned to be used as a sensor fusion assisted vision platform, and is only meant to achieve L3, and is still essentially not available to customers yet. And their L5 solution with EyeQ5 is not meant to be available until 2020 or later.
Nothing about what Tesla is doing is insane. It just seems that AP2's release was maybe 4 years ahead of industry consensus regarding when it should be achievable.
Not quite because EYEQ4 can enable L5 driving. Many companies already use it in their current prototype. the only difference between 4 and 5 is the astronomical overhead power. Power that is enough to run other companies own inefficient, power hungry algorithms.
EYEQ4 has overhead power, but not enough. 4x-5x eyeq4 however will provide enough for L5.
Or one eye4 and a nvidia (Xavier) or intel chip. Similar with what audi did with A8 with one eyeq3 for vision and another chip for sensor fusion and planning.