End to End AI - FSD Beta V12.2.1 (updated 02.20.24)
- Thread starter Phenomenally Frantastic
- Start date
Taylor Ogan is a known TSLAQ and BYD investor (TSLA competitor.) He tends not to disclose that. He also doesn’t disclose which version he is running(whether it is current) or whether the cameras are properly calibrated(this can degrade the tech quality until properly recalibrated.) He isn’t what I would call a regular user of the tech and makes hitpiece videos for youtube ad dollars. Be skeptical of the content and takes you get from him and his ilk, they are a hateful lot.
I'm excited for the tech to advance and become reliable. I am just curious how big of issues things like poor road markings or signage... or conditions like snow/slush or even mud on the sensors/cameras. This has to be an issue.
Update, it’s been since April 6 last i updated this thread. Wider release coming soon after this update to all 95+ safety scores. That’s going to be lot of people. tens of thousands more if not couple hundred thousand more.
I just wonder what in the world the occupants were doing that they didn’t hit the brakes, try to turn or anything. This was a large stationary object and yet they hit it. This is different than some truck pulling out or doing something erratic. There’s no way I’d put that much faith into this software yet.
Dude....they got Tesla FSD
They can be posting selfies and watching TikTok videos - no NEED to be wasting attention span on the road/parking lot here...
Joe without being unkind you seem to be making several assumptions here.
Not all cars are delivered to new owners with FSD software, mine does not have it activated. The take rate in the US is around 15-20% I believe which means 80-85% of new Tesla car recipients know they do not have this software. Rather, it is more likely they have basic autopilot, which is essentially cruise control, with some added safety features, but does not account for all scenarios and cannot keep a car from crashing in all instances.
If a driver accidentally swerves into a lane with a semi occupying it, while the basic autopilot feature is not activated, (essentially just the driver driving freely) a crash like this can still easily happen. Can you link any crash investigation report that this vehicle was even using autopilot or FSD? I dont see it anywhere in your link.
Further, not all cars that are delivered with FSD suite (six driver assist functions) has access to the FSD “City Streets” Beta program, (this is the one software package commonly referred to as simply “FSD” and is typically associated with autonomous driving) …one needs to apply for FSD City Streets Beta software and then prove driver attentiveness and exhibit exceptionally safe driving ability to have the FSD Beta city streets package.
FSD Suite fully activated:
Navigate on Autopilot
Auto Lane Change
Autopark
Summon
Smart Summon
Traffic Light and Stop Sign Control
Not currently activated in FSD suite:
“Autosteer on city streets”
^this is the “FSD Beta” program that everyone commonly associates with autonomous driving.
Lastly, at all times while using basic autopilot, navigate on autopilot or the City Streets beta software the >driver< is still ultimately responsible for safe operation of the vehicle. Every warning and communication from Tesla makes this abundantly clear to the operator.
It seems as if you don’t understand the features or the associated responsibility of the driver.
American drivers crash into or with semis every day in this country, Tesla is actively creating technology to prevent this from happening. Why are you so opposed to this?
You are literally listing many of the reasons that the software is NOT ready for consumer release.
Tesla is literally conducting Beta Tests using it's customers.
With deadly results.
Drink a cup of coffee and read my post again. You seem to still be deeply confused.
Crashes by customers driving their Tesla on FSD city streets beta software seems to be minimal (I’m only aware of a handful or so) and much safer per mile than the average human driver in the US.
Neither the facts nor logic are on your side here Joe.
I don't think they were using autopilot. They were definitely flying and didn't see the truck until it was coming through their windshield.
Which is, AGAIN, why software like this is NOT released to the public before it's fully tested...
If this were true, why is cruise control allowed?
Using your logic, allowing drivers to use cruise control is extremely dangerous. We all know that thousands of human drivers crash every year while using this driving feature. Should it be banned?
Lets see if you can stay consistent.
I think you are likely correct but do not know for sure either way without more information.
Using this illogical reasoning, if some authority declared FSD city streets fully tested, thousands of crash tests per year would suddenly become acceptable to you. I thought you were smart Joe.
Why are thousands of crashes and driving deaths per year on cruise control acceptable to you?
"streets" don't get "fully tested"
Software does.
And it gets evaluated against any and every unexpected condition, generally BEFORE it is launched to the general public.
Take it up with the NHTSA. They obviously have all the data and have deemed the software use by attentive drivers no more dangerous than an average driver. Why do you, in your deep ignorance, believe you know better than them?
Automated Vehicles for Safety | NHTSA
Get info on automated driving systems, also referred to as automated vehicles and "self-driving" cars, and learn about their safety potential.
THE TOPIC
The Evolution of Automated Safety Technologies
RELATED TOPIC
DRIVER ASSISTANCE TECHNOLOGIES
“Many vehicles on the road today have driver assistance technologies, which help to save lives and prevent injuries on our nation's roads. While some driver assistance technologies are designed to warn you if you’re at risk of an impending crash, others are designed to take action to avoid a crash.
The continuing evolution of automotive technology, including driver assistance technologies and automated driving systems, aim to deliver even greater safety benefits.”
Driver Assistance Technologies | NHTSA
Questions answered about adaptive cruise control, backup camera and other car tech, and videos from YouTuber’s ‘Engineering Explained’ Jason Fenske.
Because it is safe and in continual development and improvement. The very fact that it is in real world use by attentive drivers is in itself improving the technology.
Why would you want to slow the rate of improvement of life saving technology?
Info is coming. Soon they will have all the data. Including the last 10 seconds of that internal camera.
Wrong
It's a "beta" version, because it's not fully evaluated.
iOS and Windows have "been in continual development and improvement" for well north of a decade or more.
Great example, viruses that infected computers were allowed on Windows for decades. Windows software was never banned until it was perfectly secure. Thanks for making my point clear in terms that you can understand.
Now you're off in the weeds talking about "security"?
The discussion is about BASIC FUNCTION.
A Jeff Bezos hit piece likely filled with inaccuracies and omissions. 🥱
Try again slow Joe.
Uh huh.
The only "hit piece" here was the pieces of the driver littered all over the roadway in that Tweet photo.
Here were the FSD Beta V11.3 Notes from 2.19.23.
Here are the Full Self-Driving Beta V11.3 release notes in text form for improved readability.
Key points: FSD on Highways, improved lane changes, resolves recall concerns, improves AEB, report autopilot mistakes via voice notes.
More details are below, by clicking “Read more”...
1️⃣ 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.
2️⃣ Added voice drive-notes. After an intervention, you can now send Tesla an anonymous voice message describing your experience to help improve Autopilot.
3️⃣ 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.
4️⃣ 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.
5️⃣ Added a long-range highway lanes network to enable earlier response to blocked lanes and high curvature.
6️⃣ 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.
7️⃣ Improved occupancy network detections by oversampling on
180K challenging videos including rain reflections, road debris, and high curvature.
8️⃣ 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.
9️⃣ 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%.
1️⃣0️⃣ Improved overall geometry and stability of lane predictions by updating the "lane guidance" module representation with information relevant to predicting crossing and oncoming lanes.
1️⃣1️⃣ Improved handling through high speed and high curvature scenarios by offsetting towards inner lane lines.
1️⃣2️⃣ 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 [...]
Here are the Full Self-Driving Beta V11.3 release notes in text form for improved readability.
Key points: FSD on Highways, improved lane changes, resolves recall concerns, improves AEB, report autopilot mistakes via voice notes.
More details are below, by clicking “Read more”...
1️⃣ 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.
2️⃣ Added voice drive-notes. After an intervention, you can now send Tesla an anonymous voice message describing your experience to help improve Autopilot.
3️⃣ 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.
4️⃣ 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.
5️⃣ Added a long-range highway lanes network to enable earlier response to blocked lanes and high curvature.
6️⃣ 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.
7️⃣ Improved occupancy network detections by oversampling on
180K challenging videos including rain reflections, road debris, and high curvature.
8️⃣ 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.
9️⃣ 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%.
1️⃣0️⃣ Improved overall geometry and stability of lane predictions by updating the "lane guidance" module representation with information relevant to predicting crossing and oncoming lanes.
1️⃣1️⃣ Improved handling through high speed and high curvature scenarios by offsetting towards inner lane lines.
1️⃣2️⃣ 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 [...]
