A lot of people think the reason I hate LiDAR is

A lot of people think the reason I hate LiDAR is because it's expensive and it's not scalable. Partly agree, but that's not the main reason I hate LiDAR so much.

The main reason is that people who rely heavily on LiDAR believe that they are not serious about using AI.

The basis of autonomous AI is to have vast amounts of human-generated data. Purely synthesized data is not enough to train AI to learn human preferences, ethics, and values.

Self-driving AI works only if there is large-scale real data with human-generated inputs and outputs, and AI can find patterns in inputs/outputs and mimic human decision-making.

The problem with LiDAR is that humans do not use point clouds (colorblindness, low resolution, low frame rate) to make driving decisions. Humans look at colors, textures and use this information to make complex inferences.

Human driving decisions cannot be explained by LiDAR alone. If LiDAR data is used to predict what humans will do, training will simply fail.

Adding LiDAR inputs to visual inputs lowers the signal-to-noise ratio, and LiDAR becomes nothing more than a meaningless input that confuses neural networks rather than benefits them.

So as soon as I saw someone using LiDAR, I knew they were mostly just hard-coded machines running heuristics, and that their introduction of AI would be very limited and unsightly.