Search results

I am not claiming the stat is 100% accurate. But since Tesla does not release official disengagement data, teslafsdtracker is the best we got. And it at least gives us a ball park. And I agree that people may be over reporting issues as safety disengagements when they are not. So what do you...

It is based on how people report the disengagement to the website.

Safety interventions are disengagements when you believed there was a safety risk, like when you thought FSD might cause an accident. For example, it looks like FSD might hit another car or might run over an object on the road etc... Disengagements due to navigation, routing, poor lane choice...

I know this is just PR. I wonder if this change in messaging is a result of the viral incidents we saw lately. If you read the other tweets in the threads, Waymo basically says "safety is not enough, we want the community to trust us". It feels like a subtle way of saying "we are aware of the...

AFAIK, in CA, when testing with a safety driver, AV companies have to report disengagements and collisions. When testing without a safety driver, AV companies only have to report collisions. When deployed driverless in a commercial service, they only have to report collisions.

Waymo is doing 50,000 paid trips every week across three major cities now.

I got to disagree there. You don't need level 5 to do robotaxis. We already have robotaxis on public roads like Waymo and they are level 4. Remember you still need to "solve" self-driving to 99.9999% to do level 4. Both level 4 and level 5 require 99.9999% reliability. The only difference...

I will be happy when FSD is officially hands-free. There are definitely long stretches where I could go hands-free because FSD handles things well. My concern is more with the sudden "take over immediately" alerts that can be super jarring. And if there are no nags, and we are able to go...

I interpreted the lol emoji to refer to a bunch of Teslas getting stuck in the same place. I think Elon was saying that having a bunch of Teslas getting stuck in the same place would be funny. I don't think Elon was saying that notifying the fleet is a joke.

One possibility is that Elon is simply lying and Tesla does still need lidar for ground truth validation but he does not want to admit it because it would undermine his narrative about how far ahead Tesla is with the end-to-end approach. Another possibility is that Tesla is planning to add lidar...

No, Elon says that Tesla does not need to do ground truth validation anymore now that they have switched to V12. That is because end to end does not use occupancy networks. End to end also does not have a separate perception stack that requires validation. End to end takes in pixels and directly...

That is super cool to get a glimpse of the behind the scenes. It is also sort of "magical" to see these fully driverless cars come back to the depot when they are done, wait and then drive off when summoned again to pick up a rider. It seems like most of that is automated.

Well, for one, L4 cars like Waymo don't "freak out" with warnings to the driver to take over since they need to operate without a human driver. They are trained to know how to handle the situation on their own, maybe by slowing down to regain traction. It is probably one reason why Waymo spent a...

Impressive video of Waymo responding to hand gestures from traffic officer fully autonomously (no remote assistance confirmed by Waymo).

Yes, I think so. See the Mobileye imaging radar that fits nicely in the ID.Buzz AV.

This is actually the first time it has happened to me, which is why it was particularly jarring and scary. If it did happen all the time, I would be ready for it. The fact that it was the first time that it happened, took me by surprise. I think the only difference is that it had rained so the...

Yes, I think he is assuming that there will be improvements to the software both for camera vision and radar that will make camera+radar perception good enough to support L3+ autonomous driving. And it is a logical assumption because computer software is improving all the time. We have already...

That is a pretty crazy move. Waymo performed it safely but I am not sure it was wise though. Ideally, I think you want to drive in a predictable way for other drivers. So if you start to make a right turn, you should complete the right turn as expected, not suddenly make a left turn in the...

250 vehicles would be enough for a small robotaxi fleet. To put things in context, that is about the number of robotaxis that Waymo has in SF now. If these lidar are for the Tesla robotaxi, I think that would make sense. Tesla could be planning an initial production of ~250 robotaxis for the...

Yes, I think the increase in the number of fails is just statistical because we are seeing more miles. Waymo is doing millions of driverless miles now. It is inevitable that there would be some fails in that number of miles. Also, as I mentioned before, since Waymo is driverless that means there...

What if these lidar are for the upcoming robotaxi vehicle to be unveiled on 8/8? Could Elon announce something like "we solved vision-only but to make regulators happy, the robotaxi will add lidar in order to provide additional redundancy for extra safety"?

I really hate the sudden, out-of-the-blue, red hands on wheel alert. They are just so jarring. You can be cruising along just fine and then suddenly the car is blaring at you to grab the wheel immediately. On my drive to work this morning I got 2 of them back to back that scared the crap out of...

Honestly, I am not really sure how this would be any different from using the fleet to crowdsource maps. I mean if every Tesla is going to do this for their own personal routes, why not just upload it to the cloud and have crowdsourced maps that everybody can use? You even mention that people...

Motional delays the launch of a commercial driverless service to 2026 amid restructuring and layoffs: https://techcrunch.com/2024/05/07/motional-delays-commercial-robotaxi-plans-amid-restructuring/

True but you do realize Tesla likely gets hundred of thousands of interventions per day, not all of them are safety critical. It is still a lot of feedback to sift through to find the critical interventions from the not critical.

That's my point. Yes, they have more rare cases in their data but Tesla still needs to curate all that data to find the rare piece of data that helps the training the best. It is a lot of data to sift through. It is the needle in the haystack problem. Imagine if I gave you a million books and...

I think Tesla's data advantage is mostly in speeding up the first 90%. That is because Tesla can dump a ton of data that have a lot of common cases and quickly achieve ok driving everywhere in say 90% of cases pretty easily. I think we saw with the initial release of V12. Tesla was able to train...

Yes that is what he is saying. And it is why I always pushed back against the argument that Tesla would automatically win because they have billions of miles of data. Most of the billions of miles are useless. You can have other AV companies that have much less overall data than Tesla but they...

I do find these traffic violations (driving in bike lane and driving in bus only lane) to be a bit odd since we might assume that HD maps would automatically prevent these issues by labeling those lanes as off limits. So either the HD maps were wrong (the lanes were not correctly labeled in...

Yes, that is why I said that no OEM will license current FSD because liability is too high. OEMs need assurances of safety. They are not going to license a vision-only system just on Elon's word that it is safer than humans.

Yes, I saw that. I do like that Wayve has spelled out specific product types that they plan to sell to OEMs. It is a positive step towards commercialization. And yes, I think they aim to compete with Mobileye. It seems they plan to offer a very similar product and they plan to sell to OEMs...

Wayve raises $1B with contributions from Nvidia and Microsoft.

Yeah that is very bad. Deliberately driving the wrong way in a lane, especially with traffic, should never happen. And mistakes like this give ammo to the critics who say that AI is not ready to drive a car yet. I find it frustrating that AI can seem so good sometimes at driving and at other...

Cruise has mapped 40% of Phoenix: https://gmauthority.com/blog/2024/05/gms-cruise-mapped-40-percent-of-phoenix-metro-area-so-far-video/amp/

Waymo expansion!

Not autonomous driving but interesting news. NHTSA is now mandating AEB in all passenger cars and light trucks by Sept 2029. Key Highlights Mandatory AEB systems in new passenger cars and light trucks by 2029 Projected to save at least 360 lives and prevent around 24,000 injuries each year...

No, you misunderstand. I was just giving one tool that would help detect a true stop sign. It would not be the only way. Lidar can distinguish the difference between a construction worker holding a stop sign. And camera vision would also be used to detect true stop signs. So in your scenario...

The new auto park looks great. I love the visualization and being able to select the parking spot you want now. And the car executed the parking maneuver very smoothly IMO. The disengagement with the oncoming vehicle, not so great.

I found a link that works. https://weibo.com/7676370502/ObQvLf60d The disengagement with the cyclist was indeed very bad.

The link is not working for me.

I think that is why computer vision needs to be more detailed. Human vision sees the world in super high def and we see every little detail. For example, we don't just see a pedestrian, we see every detail of the pedestrian, from the clothes they wear, to the brand on their baseball cap, the...

Also, since the end-to-end model takes in vision input and outputs control, I don't think the hallucinations are likely to be perception. So for example, I don't think FSD will hallucinate an entire bus that is not really there. More likely, the hallucinations will be prediction-planning. If you...

Well, it all depends on how reliable Tesla is able to get their end-to-end model. Maybe they do get the hallucinations to be so rare that it does not matter? Or maybe Tesla discovers a way to eliminate AI hallucinations? A lot of work is going into trying to find a way to solve the problem of AI...

My understanding is that yes, FSD can "hallucinate", just like any large model. I am sure Tesla does their best to try to minimize hallucinations with quality data and the right training but the risk always there. One question is whether the hallucination is safety critical or not. If it is not...

Obviously, we want AVs to do it right as often as possible, because we want them to be the best drivers possible, better than humans. But the fact is that AVs will never do it right 100% of the time. That is not realistic because driving environments can be complex and even chaotic with factors...