Ok, I'll think I'll count this as a win then. Thanks for playing! And good luck with that robotaxi next year!
ps. I'll add "clearly lacking in knowledge and experience" in my signature, if you don't mind, so people know what to expect.
Last edited:
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Autonomous Car Progress
- Thread starter stopcrazypp
- Start date
-
- Tags
- Autonomous Vehicles
ps. I'll add "clearly lacking in knowledge and experience" in my signature, if you don't mind, so people know what to expect.
zoomer0056
Active Member
Wow, so not autonomous. Follow up question: How do you operate FSDb "city streets" with a yoke in a city environment with hands on the wheel at all times? Surely there must be a video of this, but I struggle to find one? (this is sarcasm)
Last edited:
Had to snip some posts. Just a reminder to be polite. If your post has the words “fart”, “troll” or “word salad”, please take a few deep breaths before posting. As for everyone else, thank you for your civility.
powertoold
Active Member
This is like a comedy to me.
I made a statement that I'm sure FSDb can do 1000 trips of a single route in SF from 12am to 5am.
Your response has nothing to do with SF or 12am to 5am.
I never even specified how long or easy the route is. If you've used or watched videos of 11.4.2, you'd know that it can reliably drive around, follow the law, and not hit any cars or pedestrians or obstructions, especially during 12am to 5am and especially on a prespecified easy route.
1,000 trips without disengagement (or even intervention) of an easy route in SF from 12am to 5am is doable.
Hey, me too! We found some common ground!
I presented statistical proof that what you're saying is highly unlikely to be true.
Dude, my Tesla wouldn't even complete Autopark 1000 times in a row without failure. This is just a fact. Your claim is ridiculous to me.
powertoold
Active Member
This is what I mean by logic word salad. You're all over the place. Autopark has nothing to do with fsdb. Like I said, I think you're too new for this line of discussion.
Really? You think they just bolt it in? It takes purposeful design and infrastructure to have remote control functionality added to get to L4. No one has ever provided any indication that that was a possibility in FSD. Tesla has never discussed building out that infrastructure or testing those capabilities. Elon is dead set on building an L5 vision only system to the point I think he looks like an idiot when they don’t get there - even if with a more capable computer and expanded sensor suite.
In the meantime, Mercedes, Cadillac, BMW and others will likely have capable L3 autonomous systems on highways in a year or so leaving Tesla behind. The pursuit of L5 in general is a fools errand in a consumer car, and L4 or L5 will never happen on the current sensor suite, IMO.
Yes of course I know that Autopark and FSDb is not the same thing (yet). It was an sarcastic example of a minimal ODD. *rolls eyes* Do you think it's easier or harder to implement a parking assist system or a robotaxi capable system? If Tesla cannot succeed with the easier one, why should they succeed with the other.
If you think a bit, how come Tesla cannot do hardly anything in software right outside of the core platform? Everything is half-baked and gimmicky. Like auto high beams, auto wipers, auto parking, smart summon and so on. Yet you believe that suddenly they will, in spite of years of this engineering culture, produce a robotaxi which requires "pace-maker" levels of reliability.
I'm sorry, but no one that knows anything about this area believes that Tesla will get to L4 on HW4 in the coming three years. Or this decade even. I'm 90% confident it will be never.
Last edited:
powertoold
Active Member
I appreciate your willingness to discuss these things, but this is branching off into too many side topics, so we can just revisit the reliability and CV discussion in 6 months.
Sounds great!
Let's summarise what I've said re camera-only:
1. Guessing range from a 2D image with only semantic cues is not safe enough, especially when there are few reference objects and at night. Tesla's been running into things (first responders, motor cyclists) at night. It is safer to physically measure the distance (using lidar/radar), so why wouldn't you? Some camera only solutions have cameras in the a-pillar for triangulation to improve safety. Tesla does not.
2. Cameras cannot see though dark, fog, smoke and they are easily blinded by low sun and oncoming traffic. There are too many failure modes for cameras for them to be safe enough to trust with your (or others) life at higher speed.
3. Sensor hardware is dropping in price like a rock. It's a stupid strategy to rely on too few and budget cheap ass cameras only. A front facing Lidar or Radar would save lives based on points 1 & 2 above. But Tesla is not adding that to the 3/Y. Apparently its needed for FSD only in the S/X... (sarcasm again).
Last edited:
willow_hiller
Well-Known Member
I think you need to provide proof of this claim. Every instance of hitting an emergency vehicle I've ever read about was a legacy Model S, and would have been running Mobileye code with radar fusion.
Please provide proof that Tesla Vision hit a motorcycle or emergency vehicle.
High-resolution 4D radar is definitely in Tesla’s future. Maybe HW4.5. That will help a lot, I imagine. Add a couple of side facing long range radar in the front bumper for crossing traffic and your in business, IMO.
If you put a 12 year old behind the wheel on the lap of an adult and tell the 12 year old to drive whilr the adult also has their hands on the wheel supervising and taking over when needed. Look the accident rate is better just having an adult. We must now pass a law saying everyone needs a 12 years old on their lap to drive because the 12 year old kids are so good at driving!
Last edited:
willow_hiller
Well-Known Member
You're doing the thing again where you hold Tesla's ADAS to autonomous standards. Tesla is the only OEM I know that systematically tracks collisions while their ADAS is active.
Unless you can link better safety reporting for Blue Cruise, Super Cruise, ProPilot, and Drive Pilot, I think it's fair to say Tesla has the best reporting out of any OEM.
Heck, Mercedes Drive Pilot is level 3, so surely they'd collect data on collisions and publish it to reassure their drivers they can hand the driving task over to the car, right? So how many accidents per million miles do they report?
I'm holding Tesla to the standard that they and you claim they are. What you want on the other hand is a double standard.
You and others claim that Tesla is "self driving" and that its "autonomous" and no different than Waymo/Cruise.
You and others claim that Tesla's FSD is 10 years ahead of others. I could go on and on.
You and others also claim that Waymo/Cruise are UNSAFE compared to Tesla FSD and that the hundreds of pages that Waymo has released is nothing but PR.
You and others consistently say Tesla's two sentence is THE REAL Safety Report.
You can't have it both ways, you can't claim Waymo's reporting that is 100000000x more detailed than Tesla is PR marketing.
But then uphold and regurgitate Tesla's two sentence safety report.
It has nothing to do with whether its ADAS or Autonomous.
It has everything to do with how biased someone is.
They are NOT safety reports, its complete and utter rubbish.
Its Elon showing how gullible alot of his fanbase and others are.
You do realize that under Tesla's so-called methodology, not a single accident of Waymo and Cruise would have *counted? Because airbags weren't deployed? Have you looked at Tesla crashes and seen what it takes to deploy the airbag? You can crash into a building and the airbags won't deploy. While on the other hand if i went and purposely tapped a Waymo/Cruise with my hand, they would count it as a collison/accident and have to give out detailed description of how/what happened?
Even this famous incident wouldn't count under Tesla's criteria:
It's staggering to think that even when Tesla's FSD Beta was at its worst – at launch, when it was seemingly intent on getting you into an accident/killed every other minute – they would have still boasted a 5x safety improvement based on their own methodology.
Elon in Jan 2022 said there were no accidents. Elon Musk claims zero crashes in Tesla's Full Self-Driving Beta over a year into the program
Now tell me how these two sentences is a safety report! Please take a good look at their "safety report". Compare that with competitors who meticulously document each disengagement, collision (even minor ones like tapping the car with a hand, mounting a curb, or hitting a cone), location, road type, time, detailed incident description, vehicle speed, police reports, and even simulated data.
And then you have Tesla fans using these two sentence "Safety Report" to say that Tesla is safer than Waymo/Cruise. (moderator edit)
Last edited by a moderator:
willow_hiller
Well-Known Member
You're just going to pretend that the NHTSA
Standing General Order on Crash Reporting for Level 2 Advanced Driver Assistance Systems doesn't exist?
Forget about Tesla's marketing, they're the only OEM that systematically reports crashes on ADAS to NHTSA. All the data is here for you to scrutinize: Standing General Order on Crash Reporting | NHTSA
Other OEMs don't even have the capability to track and report their level 2 data. If you read the NHTSA methodology, most of that data for other automakers comes from driver-filed complaints and police reports.
I'm pretty sure there are plenty of non-AP1 cars that have been hitting things at night (see greentheonly's twitter for example), but that's besides the point.
I wasn't discussing individual vendors or implementation - I was using Tesla as an example of a camera-only system. I was pointing out that camera-only has many failure modes and it would be a whole lot safer, especially in non-optimal situations, to have more modalities and redundancy.
Computer vision uses semantic cues to estimate distance, and if there are few cues the estimates can be off or completely wrong. You can also use triangulation, but Tesla's forward cameras are a bit too close for that to make sense. They can (and probably do) use some tricks with video and time difference to make it a bit better. I know this is used for the side repeater cameras for vision only parking.
With and active sensor you typically never misses there objects (That's called 100% recall). Lidar and radar systems doesn't need to rely on neural networks to get the distance right.
Cameras are great, more cameras are even better since it minimizes the risk of them being blinded or blocked at the same time. But there are situations where ML + Cameras are weak and other types of sensors shine.
I believe I've provided enough "proof" for people to understand this by now. I don't know about you, but I would feel a whole lot saver with more redundancy and would easily pay 2-3 kUSD extra for that to keep me and the family extra safe. I live up north so moose, and darkness and stuff. 8 months of the year.
Last edited:
No I'm going by the safety report used by Tesla and all Tesla fans. The one regurgitated every single day on here, reddit, twitter and youtube. The one with thousands of news articles written about. The one that has been read over hundreds of millions of times.
When you actually use the Tesla methodology, none of the Waymo/Cruise so called accident (which people here are losing their minds over) would count.
According to Tesla's own methodology, Waymo/Cruise HAS NO ACCIDENTS.
Which part of that don't you understand?
Waymo/Cruise would count this as an accident, Tesla doesn't. This also includes hitting the curb, cones, objects, also rear-ending cars, being rear ended, etc. Unless its a VERY SERIOUS accident, Tesla wouldn't count it as an accident. Waymo/Cruise on the other hand considers any contact even at 0mph as an accident.
Don't you see how flawed and utterly ridiculous Tesla's "safety report" is?
First admit that Tesla's two sentence is not a "safety" report but a PR report or explain how it is a safety report when less than 1% of all accidents count?
Here is another accident. This one actually got reported to insurance and police so automatically gets counted toward's NHTSA human driver statistics. But guess who doesn't count it? That's right TESLA! So the accident goes to NHTSA ~500k statistics. But Tesla doesn't count it towards theirs. Then Tesla compares themselves to the NHTSA stats and say... AHA see we are 5x better than human drivers!
When actually its just them not counting 99% of the accidents that NHTSA is counting.
Think about it, lets say you have 10 accidents. Rear-ends, side swipes, etc.
Police gets called, insurance are exchanged. It gets automatically counted towards NHTSA 500k miles stats. But it doesn't meet Tesla's airbag criteria. Now Tesla's stats is automatically 10x better than NHTSA. By virtue of just not including those accidents.
How is it that you can't see this?
Waymo/Cruise counts 100% of all collision as accidents. Tesla counts less than 1% of all their collisions as accidents. That's all that matters when comparing to Tesla.
Similar threads
- Article
