Tesla "Full Self Driving" not capable of autonomous driving

[Irace86.2.0]

FSD Beta 11.3 Release Notes


— Enabled FSD Beta on highway. This unifies the vision and planning stack on and off-highway and replaces the legacy highway stack, which is over four years old. The legacy highway stack still relies on several single-camera and single-frame networks, and was setup to handle simple lane-specific maneuvers. FSD Beta’s multi-camera video networks and next-gen planner, that allows for more complex agent interactions with less reliance on lanes, make way for adding more intelligent behaviors, smoother control and better decision making.

— Added voice drive-notes. After an intervention, you can now send Tesla an anonymous voice message describing your experience to help improve Autopilot.

— Expanded Automatic Emergency Braking (AEB) to handle vehicles that cross ego’s path. This includes cases where other vehicles run their red light or turn across ego’s path, stealing the right-of-way.

— Replay of previous collisions of this type suggests that 49% of the events would be mitigated by the new behavior. This improvement is now active in both manual driving and autopilot operation.

— Improved autopilot reaction time to red light runners and stop sign runners by 500ms, by increased reliance on object’s instantaneous kinematics along with trajectory estimates.

— Added a long-range highway lanes network to enable earlier response to blocked lanes and high curvature.

— Reduced goal pose prediction error for candidate trajectory neural network by 40% and reduced runtime by 3X. This was achieved by improving the dataset using heavier and more robust offline optimization, increasing the size of this improved dataset by 4X, and implementing a better architecture and feature space.

— Improved occupancy network detections by oversampling on 180K challenging videos including rain reflections, road debris, and high curvature.

— Improved recall for close-by cut-in cases by 20% by adding 40k autolabeled fleet clips of this scenario to the dataset. Also improved handling of cut-in cases by improved modeling of their motion into ego’s lane, leveraging the same for smoother lateral and longitudinal control for cut-in objects.

— Added “lane guidance module and perceptual loss to the Road Edges and Lines network, improving the absolute recall of lines by 6% and the absolute recall of road edges by 7%.

— Improved overall geometry and stability of lane predictions by updating the “lane guidance” module representation with information relevant to predicting crossing and oncoming lanes.

— Improved handling through high speed and high curvature scenarios by offsetting towards inner lane lines.

— Improved lane changes, including: earlier detection and handling for simultaneous lane changes, better gap selection when approaching deadlines, better integration between speed-based and nav-based lane change decisions and more differentiation between the FSD driving profiles with respect to speed lane changes.

— Improved longitudinal control response smoothness when following lead vehicles by better modeling the possible effect of lead vehicles’ brake lights on their future speed profiles.

—Improved detection of rare objects by 18% and reduced the depth error to large trucks by 9%, primarily from migrating to more densely supervised autolabeled datasets.

— Improved semantic detections for school busses by 12% and vehicles transitioning from stationary-to-driving by 15%. This was achieved by improving dataset label accuracy and increasing dataset size by 5%.

— Improved decision making at crosswalks by leveraging neural network based ego trajectory estimation in place of approximated kinematic models.

— Improved reliability and smoothness of merge control, by deprecating legacy merge region tasks in favor of merge topologies derived from vector lanes.

— Unlocked longer fleet telemetry clips (by up to 26%) by balancing compressed IPC buffers and optimized write scheduling across twin SOCs

[Dadhawk]

@Irace86.2.0 I said nothing about a mandate that the system be active, nor did I mean to say anything about the government being involved in it. All I ever intended to say was Tesla, in their roundabout way, basically has admitted that the FSD system has flaws they previously were trying to dance around with legalize.

In the end, NYFSD is a better name for it, but the name doesn't matter, even to those of us with "half a brain". The system is flawed, and those flaws encourage bad behavior as has been demonstrated by some. Those flaws endanger people that had nothing to do with clicking through a user agreement. Even the recall clearly points that out by saying what doesn't always work.

If you can't see or agree to that, then OK.

[Spuds]

It would be easier to have an actual discussion if each reply wasn't a wall-o-text...

"I didn't have time to write you a short letter, so I wrote you a long one."
-Mark Twain (and a whole bunch of others)

[Irace86.2.0]

Originally Posted by Dadhawk

@Irace86.2.0 I said nothing about a mandate that the system be active, nor did I mean to say anything about the government being involved in it. All I ever intended to say was Tesla, in their roundabout way, basically has admitted that the FSD system has flaws the previously were trying to dance around with legalize.

In the end, NYFSD is a better name for it, but the name doesn't matter, even to those of us with "half a brain". The system is flawed, and those flaws encourage bad behavior as has been demonstrated by some. Those flaws endanger people that had nothing to do with clicking through a user agreement. Even the recall clearly points that out by saying what doesn't always work.

If you can't see or agree to that, then OK.

Maybe this is a failure of my memory, but when did Tesla ever say the FSD beta software was free of flaws or had a perfect record or said the system could be used to drive hands free all the time, everywhere? In fact, they have described their system multiple times as Level 2 autonomy. The software is in beta too, as repeatedly stated.

Does having the capacity of something mean it is flawed or is encouraging bad behavior? I go back to the Hellcat with 707hp. The car's power can clearly be abused and used inappropriately on public roads. When people do that, do you blame Dodge for creating this vehicle? There are videos all over the internet of people running from the cops in Hellcats, doing donuts in intersections and gas stations. Do we blame Dodge? Do we blame manufactures for creating turbo diesel vehicles because people illegally modify the cars and do coal rolling? We don't because it is about personal responsibility to use the equipment as the manufacture intended.

The FSD beta system is a level 2 system that isn't perfect in every scenario, but should have an engaged driver capable of intervening at all times. The system might be safer in 99% of situations and less safe than 1% of cases, but that doesn't mean the system shouldn't be used as intended. Maybe it is 70/30 or 90/10. I don't know. The safety statistics show the car reduces deaths and accidents, so while it is always going to have flaws and is ever improving, it is clearly better to have it and use it as intended than not have it. For those that intentionally abuse the system, well, those are the minority and not the fault of the manufacture. I don't place blame on Tesla like I don't place blame on Dodge. I don't put responsibility on Tesla to limit the abuse of their products like I don't expect Dodge to limit the performance of their vehicles on the streets.

What was actually changed in the coding of the software for this recall? Here is another recall about the "rolling stop" function. Was the car doing anything dangerous, or was the software doing what everyone does, which is do a rolling stop or creep when coming to a stop in most situations when the driver can proceed, which is technically illegal, so Tesla had to change the software? Maybe the software would never come to a complete stop, even when it had to and would proceed, stealing the right of way or something, or coast into the intersection. I don't know. These recalls don't have much details. How was Tesla able to fix the newest recall so quickly? Did it just push a software update to temporarily cancel FSD, or did it patch the problem, or did it add more driver monitoring/driving requirements in some situations? I don't know. Do you? Because the severity of the recall could be better determined by the nature of the fix and maybe not so much in the vague language.

[Irace86.2.0]

Originally Posted by Spuds

It would be easier to have an actual discussion if each reply wasn't a wall-o-text...

"I didn't have time to write you a short letter, so I wrote you a long one."
-Mark Twain (and a whole bunch of others)

You must be reading on a smart phone. Paragraphs look different on a computer with a large screen.

If I had to boil everything down to a few arguments...

--Some people have a problem with the name FSD beta, so Musk should just rename it to something else to appease the critics and get people off his back about the name.
--Despite saving lives and reducing accidents, the system doesn't perform perfectly all the time in all situations, so therefore some feel like it shouldn't exit, which is ridiculous; this is a failure to assess risk versus reward.
--The system can be abused because Tesla doesn't do enough to prevent people from circumventing the monitoring systems, so while people don't have a problem with other manufactures failing to regulate how people use their products, they are special pleading the case that Tesla needs to regulate the abuse of these systems better.

[Spuds]

Originally Posted by Irace86.2.0

You must be reading on a smart phone. Paragraphs look different on a computer with a large screen.

If I had to boil everything down to a few arguments...

--Some people have a problem with the name FSD beta, so Musk should just rename it to something else to appease the critics and get people off his back about the name.
--Despite saving lives and reducing accidents, the system doesn't perform perfectly all the time in all situations, so therefore some feel like it shouldn't exit, which is ridiculous; this is a failure to assess risk versus reward.
--The system can be abused because Tesla doesn't do enough to prevent people from circumventing the monitoring systems, so while people don't have a problem with other manufactures failing to regulate how people use their products, they are special pleading the case that Tesla needs to regulate the abuse of these systems better.

My response:
-Musk and Tesla marketed it as "Full Self Driving", showed a demo, and described what it could do, then charged customers money for it and received investment money. He can't change the name nor back out of what he declared it to be because money has changed hands on that premise. Plus it might give his ego a boo boo.
-Give me one real example of a time the system has saved someone's life and I'll match it with 2 where it has killed someone. An attentive driver is simply better than Tesla's "FSD".
-Its not abuse if the system is designed to do that function. People are just circumventing lockouts to get it to achieve what they have been told it is. And they, like you, believe it is safer than a person driving.

[spike021]

The way I see it, my perspective being from a software engineer in the "silicon valley tech industry", if you're paying me for a service, I have a responsibility for said service to work, otherwise I'm on the hook.

In many cases that'd be a financial hook, like you may get concessions of discounts of X dollars, or Y years, etc.

For something sold to consumers in the eye of safety, it should be liability for your shit.

Because at the end of the day, if it's the "algorithm" (to sum up all the things X car company does to create "self-driving") that is driving the car, not you, then the company is/should be on the hook since "they" are the ones driving it, since it is "their" "algorithm".

[Ultramaroon]

Originally Posted by spike021

The way I see it, my perspective being from a software engineer in the "silicon valley tech industry", if you're paying me for a service, I have a responsibility for said service to work, otherwise I'm on the hook.

In many cases that'd be a financial hook, like you may get concessions of discounts of X dollars, or Y years, etc.

For something sold to consumers in the eye of safety, it should be liability for your shit.

Because at the end of the day, if it's the "algorithm" (to sum up all the things X car company does to create "self-driving") that is driving the car, not you, then the company is/should be on the hook since "they" are the ones driving it, since it is "their" "algorithm".

Especially if it's closed source. Fuck that shit.

[Irace86.2.0]

Originally Posted by Spuds

My response:
-Musk and Tesla marketed it as "Full Self Driving", showed a demo, and described what it could do, then charged customers money for it and received investment money. He can't change the name nor back out of what he declared it to be because money has changed hands on that premise. Plus it might give his ego a boo boo.
-Give me one real example of a time the system has saved someone's life and I'll match it with 2 where it has killed someone. An attentive driver is simply better than Tesla's "FSD".
-Its not abuse if the system is designed to do that function. People are just circumventing lockouts to get it to achieve what they have been told it is. And they, like you, believe it is safer than a person driving.

My attempt at a brief response:

--Musk and Tesla can't simultaneously market it as FSD capable in all situations and at all times, as you suggest, and also continue to state that it would be ready in another year and then say a year later it would be ready in another year and so on. His critics can't lament him for failing to meet the deadlines he has set, while also stating that he sold this as level 5 autonomy from day one, so which is it?
--The statistics in the link I posted showed Autopilot and FSD programs were magnitudes safer than Teslas driven without the software enabled and much more than the general population (https://www.notateslaapp.com/news/11...-safety-report). It is entirely plausible that Autopilot and FSD require more engagement of the driver, which is why they are better, besides acting as multiple assistive eyes to stop the car and avoid impacts; look over your shoulder to turn in a responsible way, and you are still taking your eyes off the road in front of you and to one side of you.
--So people need to circumvent systems of security to get the car to do something they believe it is capable of doing without these systems of security? This is like circumventing the boost control on a car and cranking it up to 30psi and blowing the motor and blaming the manufacture. It makes zero sense. People aren't accidentally abusing the system. They are knowingly doing it fully informed in spite of what they know is right. This isn't Tesla's problem.

[Irace86.2.0]

Originally Posted by spike021

The way I see it, my perspective being from a software engineer in the "silicon valley tech industry", if you're paying me for a service, I have a responsibility for said service to work, otherwise I'm on the hook.

In many cases that'd be a financial hook, like you may get concessions of discounts of X dollars, or Y years, etc.

For something sold to consumers in the eye of safety, it should be liability for your shit.

Because at the end of the day, if it's the "algorithm" (to sum up all the things X car company does to create "self-driving") that is driving the car, not you, then the company is/should be on the hook since "they" are the ones driving it, since it is "their" "algorithm".

I don't disagree. If they create robo taxis with people in the back seat or in the front seat without a steering wheel then the companies are responsible for the safety of the occupants just like the human drivers and companies of taxis, buses, trains, airplanes, etc. Liability comes from incompetence or negligence or failure of systems/components/maintence. If the plane is hit by a lightening bolt then the pilot and company aren't at fault.

What I see happening is level four or five autonomous systems will either be mandated and driving won't be an option, or there will be an option to use the system or not and companies will have systems of monitoring, so they know everything. They will know if the system is engaged or not. They will know if the steering wheel was touched or not. They will have cameras, lidar, whatever, which will provide more "eye witness" testimony than the driver's testimony. If something is the fault of the system, there will be black box data to demonstrate it. This is guaranteed if the manufactures are liable.

The end if pretty clear. Like the manual transmission, driving will seem antiquated at one point. It won't even be the safety benefits. People would rather watch TV, play a video game, make TikTok videos and FaceTime their besties than drive. People would rather have a Zoom meeting, check their stocks, work on their next business proposal than drive. If everyone could have a mechanical chauffeur then the vast majority of people would choose the system. The more people who drive autonomous cars, the safer the environment would be. The cars will be more likely to predict what other autonomous cars will do more than people, and if they can talk to each other or signal each other then that would only enhance the safety further. The risk to manufactures could drop precipitously.

[MyHybridBurnsGasAndTires]

lmao the amount of tesla d1ckriding is OFF THE CHARTS


really, really embrassing shit

[Irace86.2.0]

Originally Posted by MyHybridBurnsGasAndTires

lmao the amount of tesla d1ckriding is OFF THE CHARTS


really, really embrassing shit

lmao too I was thinking the same thing, but thinking the amount of Tesla sh1t talking haters is OFF THE CHARTS. Everyone's jumped on that lemming bandwagon


Like really, really embarrassing sh1t

[Dadhawk]

Originally Posted by Irace86.2.0

lmao too I was thinking the same thing, but thinking the amount of Tesla sh1t talking haters is OFF THE CHARTS. Everyone's jumped on that lemming bandwagon

Like really, really embarrassing sh1t

I think you are taking the whole thing way too seriously.

[Lantanafrs2]

Does this mean tesla drivers actually need to stay awake? Those shuffleboard tourneys can be tiring. Maybe book a room and leave in the morning?