diplomat33
Average guy who loves autonomous vehicles
Edging into the middle section is a common maneuver that humans do all the time. It puts you closer to the other side, so that it is easier to cross the road.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Waymo
- Thread starter Daniel in SD
- Start date
Edging into the middle section is a common maneuver that humans do all the time. It puts you closer to the other side, so that it is easier to cross the road.
powertoold
Active Member
Perhaps, I'd be a little freaked out if I were in the truck. The Waymo does some weird stuff on unprotected lefts. It's inexplicable to me.
powertoold
Active Member
I disagree. It's not common for people to wait in an oncoming traffic lane for an unprotected left. Maybe it's common in people who drive unsafely.
Thing is with a regular car, you can see the driver in most cases (I guess unless you have fully tinted windows). You can judge a bit from seeing their eyes, are they looking at you, did they notice you, are they pulling forward blindly.
With an autonomous car you don't have any clues. I think that is going to be weird. Perhaps someone will put in a visual indication.
Edit: ok I guess that's already a thing. "Autonomous Vehicle Visual Signals for Pedestrians" etc.
https://media.ford.com/content/ford...isual-language-that-could-help-autonomou.html
diplomat33
Average guy who loves autonomous vehicles
Krafcik posted a pic of the Waymo I-Pace with the latest 5th gen hardware, painted in black. Looks really good IMO!
boonedocks
MS LR Blk/Blk 19”
OMG that is H I D E O U S ! ! !Krafcik posted a pic of the Waymo I-Pace with the latest 5th gen hardware, painted in black. Looks really good IMO!
DanCar
Active Member
powertoold
Active Member
Elon's been pooping on Waymo lately. I don't think it's fair since Waymo doesn't have the data or fleet size.
diplomat33
Average guy who loves autonomous vehicles
And yet for all of that supposed data and fleet size advantage, Tesla told the CA DMV that they are still testing FSD as a L2 driver assist and that FSD Beta currently can't recognize or respond reliably to static objects, road debris, emergency vehicles, construction zones, adverse weather, large uncontrolled intersections with multiple incoming ways, occlusions, complicated or adversarial vehicles in the driving path, or unmapped roads. Meanwhile, Waymo has capable L4 autonomy and safe, reliable driverless robotaxis that the public is using every day. Hmmm?!
All of which are clearly defined to each FSD Beta tester as would befit a responsible test program.
Or, actually were you just supposed to find out yourself if/when the system fails, under the helpful "wrong thing/worst time" banner.
diplomat33
Average guy who loves autonomous vehicles
That's not the point. The point is that Tesla's data and big fleet are not putting Tesla ahead of Waymo, since Tesla is way behind Waymo on FSD.
Sorry, yes my reply was not really fitting for the Waymo thread but for the FSD Beta critique.
Cheburashka
Active Member
Exactly. Which is why Tesla makes people in EAP sign an NDA, whereas Waymo does not for it's passengers.
Exactly. And Tesla is genius at marketing whereas Google blows at consumer product marketing.
Cheburashka
Active Member
Waymo has a totally different vision.
I have a friend who works for Uber and at some point they had a bunch of engineers defect (not the ones who Lewandowski brought) from Waymo partly because they were disappointed at the agressive safety measures, meaning most of the things they made never even made it into the prototype test cars.
For Waymo safety is #1. They will not put a passenger in a car unless they are extremely confident it won't cause any harm.
For Tesla marketing is #1. They will give a half baked system to the user, label it Beta, and then figure things out as issues arise.
I have a friend who works for Uber and at some point they had a bunch of engineers defect (not the ones who Lewandowski brought) from Waymo partly because they were disappointed at the agressive safety measures, meaning most of the things they made never even made it into the prototype test cars.
For Waymo safety is #1. They will not put a passenger in a car unless they are extremely confident it won't cause any harm.
For Tesla marketing is #1. They will give a half baked system to the user, label it Beta, and then figure things out as issues arise.
powertoold
Active Member
Of course it hasn't yet. Everyone is still in the process of "solving" vision and perception. Waymo included. I'm sure you saw that mini construction zone in JJRick's video. See those cars dancing around?
No one should care who's "ahead" right now, as no one has actually solved the problem yet. And as you are solving the problem, you also need a fleet to test it in real world conditions. Once Waymo deploys a new update, they need their cars to drive on it. This is a huge bottleneck for them and why their progress has been mostly linear for the last 10 years. In JJRick's most recent video, he even suggests that the roadside assistance van follows Waymos every time there's a new update in case there are problems.
diplomat33
Average guy who loves autonomous vehicles
The "dancing cars" is a minor issue. it did not impact the performance of the Waymo at all. The Waymo still detected the construction cones, trucks and equipment and handled the construction zone with no problem at all and all while driverless too.
I never said that Waymo has perfect vision and perception. But Waymo's vision/perception is clearly good enough that they can deploy driverless cars in limited areas with confidence. That proves a certain high level of reliability in Waymo's vision/perception. Tesla's vision/perception is not that reliable yet since Tesla needs to keep driver supervision for now. Surely, you can see the difference between Waymo which has vision/perception good enough for some autonomous driving that is quite reliable but with a few issues and Tesla which is still a driver assist because they have huge chunks of the OEDR not done yet?
Again, I am not saying that Tesla will never do autonomous driving. Maybe you will be proven right and Tesla's data and fleet size will allow them to leap frog over Waymo soon. I am just saying that right now, I am not seeing that. The fact is that right now, Tesla is at L2 while Waymo is at L4.
You say that nobody should care who is ahead and yet you proceed to try to argue that Waymo has hit a bottleneck and won't make progress fast enough. And Elon and some Tesla fans are constantly trying to prove that Tesla is better than Waymo. They are constantly pooping on Waymo. If Tesla has all these advantages and Waymo has no chance, then why attack Waymo and try to put them down all the time? I think Elon and some Tesla fans feel threatened about Waymo or they wouldn't try to put Waymo down all the time.
If you look back 5 years, Waymo's progress has not been linear. Waymo's autonomous miles have increased exponentially. Now, they are well over 20M.
And Waymo's disengagement rate does not look linear either. In just 5 years, it went from 1,563 miles per disengagement to almost 30,000 miles per disengagement. That's a 1916% improvement. if that's not progress, I don't know what it is.
Of course, we don't know what Tesla's disengagement rate is since they won't release it.
But I think the big difference is that Tesla is at an earlier stage of development where qualitative progress is more obvious. For example, we might see that FSD Beta almost hit a parked car 5 times in videos a month ago, but now, does not hit any parked cars. That's very visible, obvious progress. Waymo already has very reliable perception/planning so we don't see that type of obvious progress. Waymo simply does not make a lot of those types of mistakes anymore. Waymo's progress is more subtle as they gain smoother planning or smarter driving policy that makes the driving safer or more human like. Waymo's progress is measured in safety reports and as we see them deploy in more cities.
powertoold
Active Member
I’m not saying that they hit a bottleneck. I’m saying that their bottleneck has always been there. Waymo has a huge challenge when it comes to validating their software updates in real world conditions. For example, a minor change in the driving policy may work well in Chandler Arizona, but it presents a new edge case in San Francisco. Without a fleet that is constantly encountering edge cases, it’s difficult to know how your changes affect one driving locality vs another.
But they are driving in San Francisco. Surely they are sharing their experiences between all the locations?
Similar threads
- Article
