Comment by madamelic | Hacker News Reader

Comment by madamelic | Hacker News Reader

I strongly believe that we are going to see many brands hit L4 and L5 before 2030.

Multiple manufacturers seem to be circling on the same advancements and leaps, they all seem to be working with each other to get regulations in place for them to deploy their cars.

I do believe we are about to see AV's version of "making reusable boosters being reused boring" moments with AVs where suddenly multiple companies are doing what others thought impossible years before. Even the much demeaned Tesla FSD is shockingly human-like and reliable on v13.

The issue that Waymo specifically is going to have is scalability. Unlike other brands who have AV, Waymo doesn't have manufacturing capabilities in-house currently nor logistic or a generalizable model. I think Waymo is definitely ahead of the pack but time will tell if the brands who went slower will pass them up due to Waymo not having a financial reason to push to financial viability on any timeline shorter than "after Google runs out of money".

If Waymo's plan is to have to map the entire US to be capable of driving in it, that's going to cost a lot of money and take a lot of time vs having cars that can roll out of the factory and drive straight to their assigned city unassisted.

Interesting! I take the opposite of this argument.

I ride in Waymos constantly. They are boring technology to me that I no longer think about. Effectively complete trust in them for the areas they drive in. The driving is so steady and consistent that I forget I'm in an autonomous vehicle. The only thing I want is for them to be able to take me to Oakland and to SFO via the freeway, but am comfortable waiting for those to become unlocked on the assumption that my trust level will remain consistent as their unlocked region footprint grows.

I don't trust Tesla's self-driving as far as I can throw it. I'm not a huge Elon hater or anything like that. https://teslafsdtracker.com/ gives me pause! 1 in 10 rides has a critical disengagement and it hasn't improved in three years. I will concede that the distance travelled appears to be improving rapidly, and increasing distance could explain why rides continue to have a critical disengagement, but man I just can't overstate how uncomfortable that makes me feel. I want nothing to do with sitting in the back seat of an autonomous vehicle that needs someone to take over every 1 in 10 drives.

Also, as a consumer, the notion of wanting to choose a system that relies purely on vision over one that is a combination of lidar and vision is just nonsense. Just because humans drive with vision + thinking doesn't make me feel like that is the ideal solution. I want systems that use all the tech at the machine's disposal to make me as safe as possible by handling edge cases that a human driver would fail at.

For Tesla to even be in the conversation with Waymo is very promising for the concept of low-tech solutions to autonomous vehicles, which in turn is very promising for how scalable this technology can be.

As far as I know, a self-driving Tesla, such as what they are operating in Texas, is almost the same hardware as an off-the-shelf Tesla, whereas each Waymo is a custom built vehicle, with added manufacturing costs (on top of the base vehicle) in the hundreds of thousands of dollars. The significance of the Tesla service being comparable to Waymos cannot be overstated IMHO.

"If Waymo's plan is to have to map the entire US to be capable of driving in it, that's going to cost a lot of money and take a lot of time"

So Waymo doesn't have access to Google Maps/Street View vehicles' data? Because that sounds like what you're describing here.

Also, would NYC be the first real test city for self driving with seasons? Besides SF's 3 daily seasons. The city should put up bleachers and sell tickets for the first snow.

Waymo has to scan with its entire sensor suite which includes LiDAR. I am not entirely sure if the Google Maps cars has LiDAR equipped. Definitely would seem sane to do since they are both Alphabet products even if Maps has no use for the data.

I don't believe that at all. The problem is same like replacing managers with AI - no accountability. Computer can't make decisions which may or may not kill people, because when that will happen, who is responsible for injury? Driver who is not driving? Developer of such system? Company who made the car? Company who created the self driving system?

> Developer of such system? Company who made the car? Company who created the self driving system?

Those would in many cases be the same company. And yes, the entity responsible for injury would be the car manufacturer, not the driver or riders. Taking on legal liability for the car is part of the 'qualifications' for L3.

The specifics of what part of the system failed can be litigated if defendants (companies) need to be added to the case but in no case will the people inside of a vehicle designated higher than SAE L3 be held at fault for the accident.

And that's why I think we will never see L4 or L5 system. These system would be constantly "broken" due to i.e. dirty/miscalibrated sensors to prevent possible lawsuits which would be so often that they would become essentially useless.

The company who operates the car is responsible for this. Broken systems is not a way to avoid lawsuit but a sure way to be convicted.

If system is broken, it is not working. If it is not working then it can't drive you around.

If today I buy a brand new car, drive off the lot and the breaks fail causing me to plough into a pedestrian and kill them, who is to blame?

In my jurisdiction, damages from car crashes are strict liability, so you would in fact be legally liable. Limited to ~10 million USD for damages to objects, no limit for damages to persons. Of course the manufacturing defect would give you a credible claim against the manufacturer, but that's a separate matter. Which is why automotive liability insurance is mandatory.

Doesn't this answer the question then? Dodgy breaks or dodgy AI, it's on you if your car cleans out someone crossing the road.

You are for failing to check that your car is in drivable state before setting on your journey. This is actual law in my country.

The manufacturer obviously, but they can sell the car in the first place because this defect risk is quantifiable for their liability insurance provider, who can evaluate how risky said car company is in terms of their manufacturing and how likely it is they'll need to pay out a claim, etc.

For self-driving, that evaluation is almost impossible. Sure it can look good statistically, but for things like brake lines, brake pad material, brake boosters, etc, they are governed by the laws of physics which are more understandable than any self-driving algorithm.

I think with Waymo we're probably at the point where an insurer could have decent stab at what their liability would be if asked to cover AI-related accidents. In fact, given that these cars are on the road and have reportedly been in accidents, I would imagine this is past being a hypothetical concern and now well into the territory of 'solved problem'.

2030 is an interesting target date. Waymos were driving on public roads for eight or nine years before taking the safety driver out of the vehicle and offering rides to the public.

Elon, of course wants to do that much faster. But he was also supposed to be on Mars by now.

Reply on news.ycombinator.com