With reports that Tesla is going back to radar over Tesla Vision next year, what happens to those of us with new cars that don't have USS? And what about the stupid (me) folks that bought Enhanced Autopilot.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Just received my new car without ultrasonic sensors
- Thread starter C-rob
- Start date
- Status
This. I'm going to wait til this problem is resolved before ordering. This will be my first Tesla but I'm in no rush to buy so waiting isn't an issue.
I get the move away from USS, but I live in the UK in a driveway surrounded by limestone walls, the only way to park without sensors is to get a family member to help. For a new "luxury" car this is hopeless! It was not made aware that this feature would not be available, otherwise I would not have picked it up! Why hide it? Hopeless...
Eating my own words a bit though - still taking delivery of my Tesla tomorrow because my current car is falling apart and more expensive to repair than sell after 14 years.
However if you can wait, it makes a lot of sense to see where the dust settles with this radar issue and what the competition offers. Maybe it'll push Tesla to be more consumer friendly.
I still believe right now in Dec 2022, the Tesla Model Y is the best EV SUV option available at its price range. There's better internal combustion engine cars out there for sure but all the EV options at this price range have it's own product issues (e.g. Ford or Volkswagen software and/or I find the EQB just to be ugly) or customer experience issues (dealer markups or long delivery times).
This is Tesla’s logic but it falls apart when you factor in that human brain is magnitudes more sophisticated than FSD and can compensate for the lack of sensors. And that is not necessary true because lack of sensors (e.g. short/nearsightedness, missing an eye or hearing) is still noted as impediment.
In reality, fMRI scans of human brains are quite interesting because we not only subconsciously track a lot of objects but also predict their behavior - on a scale much larger than the “occupancy network”. The thing is that we even anticipate things we haven’t seen yet.
Presumably you have discovered the function that puts the camera views on screen? And overlays the path the car is going to follow when you reverse on the rear view? And you have wing mirrors?
I genuinely have trouble understanding the hyperbole on these USS threads. Removing them from orders already in progress is a dick move, to be sure, but from these posts anyone would think that the car was literally unusable.
FSD actually does this too. One of the sections of the AI day discussion talks about how the car deals with parts of its view that are occluded (other cars/trees/street furniture, etc) and it models things that could have happened in the spaces it can't see and factors that in to its planning. Timestamp And unlike a human, it should operate on a real probability model, not one based on recollections of that one time you did a thing and either it paid off/turned out badly enough to make you remember it.
I think is a combination of what was once there and still there on cars a few months older + the uncertainty on how non USS cars is to function. Customers basically feels like Beta testers without opting into becoming a Beta tester. Non USS car owners also didn't get any kind of reimbursement for missing a feature that is present on majority of similar cars so you feel as if you paid for something that you didn't get kind of thing.
How this will turn out, not sure. I am an owner of a non USS Tesla. Is it good, bad, ugly? Still not sure and I guess time will tell.
Last edited:
The point is that it is still quite primitive compared to how humans do it. Hence, the “beta” designation that will stay for quite some time.
People are generally hyperbolic and those on Internet forums even more so. I don’t have the USS sensors. I wish they had a fix before removing them, but the car is great and easy enough to park. Now the people that paid for FSD and still don’t have it, they have something to complain about. I just assumed it would never work for the lifetime of this car and was pleasantly surprised to see the TACC and AS work on almost all streets; I thought they were limited to highways.
I agree, I wouldn't buy FSD.
I think Tesla are doing some really incredible things in their efforts to achieve FSD but whether they're going to get there any time soon (or even if they're going to get there at all - this is cutting edge stuff after all) very much remains to be seen. Given how immature EVs are at present I only plan to hold on to this one for a couple of years and then upgrade to whatever the leading edge is at that time and there's basically zero chance of FSD doing anything useful in the UK in that timeframe.
In real terms it's still basically in the lab at the moment, so how well it performs remains to be be seen, I grant you.
I'm not sure I'd call it primitive though. Humans are very good at solving novel problems and being creative, but computers have 360 degree awareness, perfect concentration and memory recall, an ability to exactly compute trajectories and velocities and the benefit of millions of real world experiences from which to determine outcome probabilities.
If you're driving along when there's a spontaneous flood, alien invasion or something that the self driving software has never encountered before, then yeah I'd bet on a human to do better than a car. If it's 'is it safe to take this gap across 6 lanes of traffic with patchy visibility' (like the model in that AI day presentation) then a computer will eventually be better at it.
Very much agree. The main issue I have is that FSD beta efforts affect the non-FSD part of the car.
What is the final goal?
To match or be better than humans in a limited set of scenarios? Yes, AI is pretty close with Tesla not being the leader there.
To match or be better than humans in the entire, “normal” set of everyday scenarios? That has not been clearly defined yet - is rain normal? Is road construction normal? Is unpaved road normal? Etc. Tesla seems to be focusing on that. Very much beta and it will stay beta for a while.
To be better than human drivers, period? That will take many, many years. Reference: Finnish rally driver
Without clearly defined goal success is also murky.
I think we're already beyond that actually.
Statistics for collisions per mile driven, vehicle repair costs per mile driven, or even deaths per mile driven, are not that difficult to come by. The entire insurance industry revolves around having these numbers, having them be current, and having them be accurate. Beyond third party metrics, Tesla themselves can uniquely collect more of this type of data than any other auto maker, and most importantly, can correlate that data according to who was driving - the human, or the car.
Once the FSD fleet statistically surpasses humans in these metrics, it's difficult to argue that success has not already been achieved.
(Perfection is a different story, and there will always be corner cases to refine and improve. But the same applies to human drivers - its why so many do poorly on black ice, over-correct on skids, drown when entering flooded roads, etc, etc - situations where, with additional training, human drivers also could have done better.)
The stats from Tesla are misleading (“winner’s bias”) because they are limited to a set of scenarios where FSD can work and there is a “safety net” - a human who has to take over when the FSD simply gives up.
We have to compare FSD against humans on the same road, under the same conditions and every FSD failure is a crash, as every human failure is a crash. Unfortunately, that data is not available.
We have to compare FSD against humans on the same road, under the same conditions and every FSD failure is a crash, as every human failure is a crash. Unfortunately, that data is not available.
Currently FSD should be used, similar to Cruise Control. Turn it on when you are on long, boring trips, but keep manual control when things get dicey.
@Boza - I understand your points, but I don't think that impacts the ability to claim success.
On limited scenarios - Let's say the comparison is Vehicle Set A (a group of cars with approx 70% FSD driving, 30% human driving in tough "non-FSD-able" conditions), vs Vehicle Set B (100% human driving in all conditions), both sets statistically large and doing statistically average driving over a period of time. Sure, we could say that FSD "skipped the hard parts" - and that's true. But if the statistics (FSD augmented by a human, vs human driving alone) say that FSD was a net safety benefit - that's still a success, even if FSD wasn't driving 100%.
And on "when the FSD simply gives up"... don't forget that humans do that too...
On limited scenarios - Let's say the comparison is Vehicle Set A (a group of cars with approx 70% FSD driving, 30% human driving in tough "non-FSD-able" conditions), vs Vehicle Set B (100% human driving in all conditions), both sets statistically large and doing statistically average driving over a period of time. Sure, we could say that FSD "skipped the hard parts" - and that's true. But if the statistics (FSD augmented by a human, vs human driving alone) say that FSD was a net safety benefit - that's still a success, even if FSD wasn't driving 100%.
And on "when the FSD simply gives up"... don't forget that humans do that too...
- Status
Similar threads
