Hacker News Reader

Comment by sefrost | Hacker News Reader

I just visited San Francisco for a few days (first time!) and it was absolutely amazing seeing all of these cars driving around everywhere.

I also thought the driverless car companies had stalled (bad pun); but that is obviously not the case. I reflected on why I thought that and I realized I had allowed news about the Tesla self driving technology to negatively affect my understanding of this whole market.

I have since watched a bunch of Waymo videos online, they are so cool! The possibilities seem huge.

View on HN

Fwiw, the outward deployment of Cruise has stalled, but it was mostly because withheld accident footage rather than anything to do with the tech. They were nipping on the heels of Waymo, and they have continued development internally while they wait for their time-out from the regulators to be over. I expect them to be Lyft to Waymo’s Uber.

I don’t know about that. Cruise tried to move too quickly in an attempt to catch up to Waymo and accumulated a ton of much publicized issues in SF. Constant stalling in intersections, involved in a few collisions, etc. CA DMV clearly noted “performance issues” as one of the reasons for their suspension.

The other incidents were much publicized, but I don’t think that had anything to do with their actual burden/costliness, which was quite tiny. After all, a Waymo car got violently totaled by a crowd even though they have had an almost flawless record. Some fraction of the population hates these things intrinsically, and I don’t think the disdain is driven by the actual risk or events; those are just an excuse.

Some of the incidents were overblown, sure, but even before the pedestrian dragging incident the CA DMV had ordered them to scale back their fleet by 50%.

I took a few rides with Cruise and it was clear it was not ready for prime time, the potential was evident but it seemed like they were a couple years behind.

I'm not disputing that Cruise was behind Waymo a bit, just the idea that development was stalling because supporters were losing faith in progress. (That was the possibility raised in the original comment I was replying to.) That is, there is no "autonomous driving winter".

Separately, any level-headed quantitative assessment of the cost-and-benefits of the roll-out to society will find that we are going way, way too slowly. It is madness to slow down a research program -- whose every day of delay costs the US over a billion dollars -- because, out of hundreds of test vehicles driving over many months, a handful blocked traffic for a few minutes. If you wait to start building out the fleet until autonomous vehicles are 2x-10x safer than human drivers (the conventional wisdom among regulators), it will delay things by years and so literally cost tens of thousands of lives.

(I've taken dozens of rides in a Cruise and had zero safety issues, am very happy to have them driving around in my neighborhood, etc.)

Is it possible you were similarly mistaken about FSD?

FSD vs Waymo is not even close.

True. FSD works anywhere while Waymo works in a few square miles.

FSD is level 4 nowhere. The human is always responsible for driving.

Tell me then, what Tesla do you own? When's the last time you used FSD? What version did you use?

And if your answer is "I don't own a Tesla" why should anyone take you seriously?

I own a Model 3 and have used FSD v12.3.x as recently as a month ago. I've also taken several Waymo rides in that time. I can confidently say it's not even in the same league as Waymo.

Maybe! I get quite a lot of Ubers here in Vancouver and they are almost exclusively Tesla. I don’t known if the “3d view” that is displayed on the screen in the dashboard is used by FSD or not; but if it is there is no way that could be safe for self driving cars. It’s so inaccurate!

I could be wrong about their progress though, always happy to be corrected.

What you see in those cars is not FSD, but Autopilot that comes with every car. Autopilot is lane keeping + traffic aware cruise control plus a few small features like green light chimes and speed sign detection.

FSD looks very different and you can see everything in the road and what the car is planning to do. There's some good YouTube channels where they test different versions of FSD in different situations, AIDRIVR does a good job, eg https://youtu.be/rMDNFLsXFEU?si=SHWAm24ZY3psnO4Z&t=136. Nice thing about those channels is you can also look at old videos and see the rate of improvement.

I had FSD for one month on my car and it was really impressive. There were plenty of disengagements, but they were all what I'd call "quality of life" issues. Car was too hesitant during blind lefts, or advanced lefts, car slowed down too much for speed bumps, it got into turn-only lanes later than I'd like, etc, etc. The hardest part of driving in my town is a 3 lane roundabout where upon exiting the roundabout you have to very quickly and confidently change 2 lanes in order to get on the highway, it handled it with ease. Overall, I would not buy it today, but didn't see any reason why they can't solve the issues I saw.

Reply on news.ycombinator.com