Hacker News Reader

Comment by maxlin | Hacker News Reader

This article will age so badly lol.

Tesla's architecture is the only one of the robotaxi providers which actually _can_ scale. Not being dependent on exact HD maps, having the same standard as humans do "if you can see enough to drive" means eventually there'll be scant roads Robotaxi can't serve. By this strength it'll serve in areas that would never work for LIDAR/pre-HD-mapping requiring solutions

View on HN

I've never understood this argument. When I drive, I'm relying on more than just vision. I'm relying on sound, tactile feedback, an understanding of traffic law, an understanding of human behavior on the road (some of which is specific to my locality, the time of day, etc.), an understanding of weather conditions and geography, past problems I've run into on the road and heuristics I've learned over time to improve safety, weighing the risks vs benefits of certain maneuvers (especially considering how specific cars around me are behaving), and so forth. Some of this a purely vision focused FSD could simulate. But much of it it cannot. That's why alternate sensing systems really are needed, whether it's LIDAR or something else.

Which of the things in your list can LIDAR provide a replacement for?

Every thing is a sensor. It provides input. Tesla decided to remove more and more input while claiming it’s better. Even Tesla knows because the robo taxis need more sensors than they sell in stores. So latest HW4 you buy is already obsolete

In the sense that LIDAR can detect some things that visible light sensors cannot, it can make up (in some respects) for the lack of those other capabilities. It's not meant to replace them per se, but to provide another dimension of detection.

Except for right now a robotaxi requires a driver

Saying Elon Musk won’t succeed at something that is a hard technological problem is foolish, yes. His companies have done what everyone said was impossible for ages.

But I do wonder if rising nations will instead just build their infrastructure to line up with technology (e.g. HD mapping their streets as they build them, installing signed beacons to inform AVs). The cost of this tech is cratering. One possible future, though admittedly an unlikely one.

I do see that as much more likely for China than western, more ossified cities.

But even then, it'd only be for cities. FSD functioning basically as a human driver would function, anywhere they would function is the real deal.

Reply on news.ycombinator.com