There is a longstanding argument in autonomous driving that goes roughly like this: the road is dangerous and unpredictable, so you should throw every sensor you can afford at it - lidar to measure distance, radar to see through weather, high-definition maps to remember what the cameras might miss. More sensors, more redundancy, more safety. It is an intuitive position, and it is expensive, and Nullmax has spent nearly a decade arguing that it is wrong.
Nullmax's position is that a car should drive the way a human does - by looking. Cameras, deep learning, and a lot of training data, but no lidar and no pre-built high-definition maps. This is not a cost-cutting compromise dressed up as a philosophy; it is closer to a genuine philosophy that happens to also cut costs. If it works, you get autonomy that is cheap enough to put in ordinary cars. If it doesn't, you have removed the safety net that everyone else insisted on keeping.
The interesting thing about Nullmax is that the technology is already in cars people buy. This is a company that ships, which in the autonomous driving world - a field with an unusually high ratio of impressive demos to actual products - is worth paying attention to.
