And it costs about 9% of what a new Model 3 would run you. It's also coming to market almost 5 years after the Tesla HW3 stack that runs almost all cars on the roads to which you're making the comparison.
> My argument on autonomous driving is that it can't just match the safety of the average human driver. It needs to be 100-1000x better, unquestionably better for everyone, always.
That sounds like maybe a marketing argument? You're saying that if it was only 2x better at some visceral metric (total traffic deaths, say) that you think people wouldn't accept it and it would be regulated out of existence, either by liability concerns or actual lawmaking?
The counter argument is that those metrics are, after all, visceral. The US has 40k traffic fatalities per year. You really think the government is (or courts are) going to kill a technology that could save 20k people every year?
No. "Merely better" is the standard that will stick.
> I think that's probably only achievable with LIDAR
Everyone on your side of the argument does, but frankly this has become an argument of faith and not substance. LIDAR capability has become a proxy for "not Tesla", so people line up along their brand loyalties. For myself, LIDAR is actually the technology looking like it isn't panning out. It's not getting cheaper or more reliable. LIDAR autonomy isn't getting better nearly as fast as vision solutions are iterating.
Most importantly, LIDAR at its best still only gives you the 3D shape of the world, and that's not enough. LIDAR doesn't help with lane selection, it doesn't help with traffic behavior modelling, it doesn't help with sign reading. It doesn't help with merge strategies. All that stuff has to be done with vision anyway.
The "don't hit stuff" bar has, quite frankly, already been crossed. Tesla's don't hit stuff either. If you give them LIDAR, they.... still won't hit stuff. But all the work, even for LIDAR vehicles, is on the camera side.