FSD Beta v11.x

The next big milestone for FSD is 11. It is a significant upgrade and fundamental changes to several parts of the FSD stack including totally new way to train the perception NN.

From AI day and Lex Fridman interview we have a good sense of what might be included.

- Object permanence both temporal and spatial
- Moving from “bag of points” to objects in NN
- Creating a 3D vector representation of the environment all in NN
- Planner optimization using NN / Monte Carlo Tree Search (MCTS)
- Change from processed images to “photon count” / raw image
- Change from single image perception to surround video
- Merging of city, highway and parking lot stacks a.k.a. Single Stack

Lex Fridman Interview of Elon. Starting with FSD related topics.

Here is a detailed explanation of Beta 11 in "layman's language" by James Douma, interview done after Lex Podcast.

Here is the AI Day explanation by in 4 parts.

Deep Understanding Tesla FSD Part 1: HydraNet

From Theory to Reality, Analyze the Evolution of Tesla Full Self-Driving

saneryee-studio.medium.com

Here is a useful blog post asking a few questions to Tesla about AI day. The useful part comes in comparison of Tesla's methods with Waymo and others (detailed papers linked).

Twelve Questions about Tesla Autopilot Technologies Exposed at 2021 AI Day

On Aug. 19 2021, Telsa held an AI day , exposing more technologies about Autopilot , from perception, planning, data labeling, simulation…

yuhuang-63908.medium.com

 

[FSDtester#1]

Not that my eyebrows aren't going up a bit as well, but it does make sense. The (two) motors run on pulse-width-modulated pulses with a ridiculous number of electronic control loops, looking at current, temperature, rpm (both of the fronts and rears), with hardware assist on the PWM end of things.

I've worked on fan motors and the design of the surrounding electronics, which aren't nearly as complex as what Tesla plays with. And it's easy to make a mistake somewhere. Some mistake putting poles in the right half plane resulting in surges.. sure, it's possible. And given manufacturing variations on silly things like resistance, inductance, capacitance, and what-all, it's believable that what the engineers played with Worked Just Fine.. but then, given big statistical runs of parts, had corner cases that would lead to instability. Which doesn't take much thought to realize would lead to vibration.

So, they hit some poor engineer with a wet noodle for a while, asking, "Just when did you take that control theory course, anyway?!?", then get a new load of software that keeps the poles in the left half plane, where they belong.

I just hope they aren't planning on reducing acceleration.
A service technician about a year ago took me on a test drive in my car.
We were already in service mode and he said "Did you know that your car has another 200 to 300 horsepower?"
I said I didn't, he then asked if he could show me, I of course said yes.
He entered a code while in service mode and then punched it going onto a freeway.
Holy smokes, it took off like never before. Hopefully they will offer a performance boost to all the Plaids in the future. Probably not, since it would likely create more warranty concerns.

 

[KArnold]

When they say "FSD is x% better than humans" do they really mean "FSD with safety driver" or FSD alone?

Tesla has played some stats fairly loosely.

 

[AlanSubie4Life]

When they say "FSD is x% better than humans" do they really mean "FSD with safety driver" or FSD alone?

Tesla has played some stats fairly loosely.

Fortunately, Elon has been unambiguous, finally providing complete clarification on Twitter recently:

This is good stuff and represents Tesla’s objective and definition (L4/L5!). We’re nearly there!

Elon: FSD Beta tweets

 

[sleepydoc]

This is not what I am talking about, obviously.

How is that not clear? If nothing else, from further context!!!

We were talking about freeway driving. This implies speed. Further context clearly indicated the scenario. Encountering events that are at the edge of perception range…etc…lead traffic helps…etc.

Can also apply to something as simple as a traffic light.

Not complicated stuff. The fix may well be impossible though!

Hmmm…Context…let’s see what context you were using:

Anecdotally, my Waymo rides with no human driver (sample size in the dozens) have been MUCH smoother than any FSD drive I’ve ever taken (sample size in the hundreds).

Of course, FSD is able to hop on the freeways and take me places Waymo still cannot. And FSD is pretty darn good on freeways now, the only annoyance being constant nags.

But when it comes to in-city drives, Waymo is hands down a lot further ahead than Tesla.

MarcG Said:
the only annoyance being constant nags.

AlanSubie4Life Said:
And having to disengage whenever encountering stopped or slowing traffic. So annoying.

Yes, it’s very clear. You specifically quoted the line where Marc was talking about freeway driving, even deleting out the rest of his post. Yep. The context was needing to disengage when encountering stopped or slowing traffic on the freeway which I pointed out is not an issue.

 

[AlanSubie4Life]

The context was needing to disengage when encountering stopped or slowing traffic on the freeway which I pointed out is not an issue.

This is indeed an issue. Plenty of examples readily produced (as mentioned, I’m not going to bother gathering and posting them).

You pointed out for: “stop and go traffic and slow-moving rush hour traffic. No issues what so ever.”

This is of course a completely different scenario than “encountering stopped or slowing traffic on the freeway!!!!”

A couple of the key differences being speed, and having no lead vehicle.

 

[sleepydoc]

Fortunately, Elon has been unambiguous, finally providing complete clarification on Twitter recently:

View attachment 987665

This is good stuff and represents Tesla’s objective and definition (L4/L5!). We’re nearly there!

Elon: FSD Beta tweets

Elon has also said in the past that FSD will need to be 10-100x safer than humans to be accepted.

He’s correct - humans are horrible bad at perceiving risk and interpreting statistics. If a hypothetical FSD system has 1/100th the number of accidents that human drivers do but the accident it does have is somethign a human would have prevented (or even potentially have prevented) the public sentiment will be that the FSD system is a dangerous failure. People see the accidents that do happen, not the ones that don’t.

 

[SSDTester#1]

When they say "FSD is x% better than humans" do they really mean "FSD with safety driver" or FSD alone?

Tesla has played some stats fairly loosely.

FSD with a safety driver watching it, while the interior camera watches the safety driver. You know, can never be too safe.

 

[jebinc]

FSD with a safety driver watching it, while the interior camera watches the safety driver. You know, can never be too safe.

Mirror + knuj

 

[JulienW]

Elon has also said in the past that FSD will need to be 10-100x safer than humans to be accepted.....

Even then it will be a tough sale since the media will focus on every accident, injury and especially death anecdotally magnifying the horror and scaring the public. Also the irony is the fewer the deaths the more they will be covered and reported over and over. Stalin was correct: A single FSD death is a tragedy but many thousands of human cased traffic deaths are simply statistics and just business as usual.

 

[AlanSubie4Life]

Even then it will be a tough sale since the media will focus on every accident, injury and especially death anecdotally magnifying the horror and scaring the public. Also the irony is the fewer the deaths the more they will be covered and reported over and over. Stalin was correct: A single FSD death is a tragedy but many thousands of human cased traffic deaths are simply statistics and just business as usual.

Fortunately we have a few years before we need to concern ourselves with this.

It’s not like autonomy is going to sneak up on people and just take over driving suddenly.

 

[sleepydoc]

Fortunately we have a few years before we need to concern ourselves with this.

It’s not like autonomy is going to sneak up on people and just take over driving suddenly.

I dunno - I heard V12 was

 

[sleepydoc]

This is indeed an issue. Plenty of examples readily produced (as mentioned, I’m not going to bother gathering and posting them).

You pointed out for: “stop and go traffic and slow-moving rush hour traffic. No issues what so ever.”

This is of course a completely different scenario than “encountering stopped or slowing traffic on the freeway!!!!”

A couple of the key differences being speed, and having no lead vehicle.

well, that’s not what you said at all But I’ll expand my statement - I’ve never had any issues with stopped, slowing or stop and go traffic either. Slowing traffic is also not at the ‘edge of perception’ as you allege. The only thing that might be considered as such is if you’re driving 70 MPH and suddenly come upon traffic that is completely stopped with no slowing cars in front of you. I can’t remember the last time that happened to me.

 

[D Good]

I just hope they aren't planning on reducing acceleration.
A service technician about a year ago took me on a test drive in my car.
We were already in service mode and he said "Did you know that your car has another 200 to 300 horsepower?"
I said I didn't, he then asked if he could show me, I of course said yes.
He entered a code while in service mode and then punched it going onto a freeway.
Holy smokes, it took off like never before. Hopefully they will offer a performance boost to all the Plaids in the future. Probably not, since it would likely create more warranty concerns.

Proof? The car is nearly traction limited so it is kind of hard to believe that they would do this/this happened. If they could make the car faster they would. Plaid is no longer the quickest road vehicle, lucid air sapphire holds that crown (and maybe the rimac nevera depending on what you count as a road car).

 

[AlanSubie4Life]

Slowing traffic is also not at the ‘edge of perception’ as you allege. The only thing that might be considered as such is if you’re driving 70 MPH and suddenly come upon traffic that is completely stopped with no slowing cars in front of you.

This is what I am talking about, it is quite common. In San Diego it is routine to have free-flowing traffic followed by common sudden slowdowns (see I-805S north of Golden Triangle south of the I-5 merge for example). It is just routine. Caused by massive amounts of added traffic at a particular location and time of day.

LA too. Any big city I would guess!

Anyway it is a trivial problem for a human. I’m not sure whether or not FSD with current sensors can do it. I don’t know how far out they are allowing the system to respond.

 

[FSDtester#1]

Proof? The car is nearly traction limited so it is kind of hard to believe that they would do this/this happened. If they could make the car faster they would. Plaid is no longer the quickest road vehicle, lucid air sapphire holds that crown (and maybe the rimac nevera depending on what you count as a road car).

I hear ya, if I would of known what was going to happen, I would of took out my phone and recorded. That technician was the head tech at the service center. I since have asked people at another service center and they didn't know about it. We also were going up a slope entering the freeway, and he hit it from about 30 mph or so, there was no slippage. So as far as proof goes, put this as something you heard about from some weird guy on the internet said happened. I was there, so only I can truly believe it.

 

[EVNow]

Fortunately, Elon has been unambiguous, finally providing complete clarification on Twitter recently:

View attachment 987665

This is good stuff and represents Tesla’s objective and definition (L4/L5!). We’re nearly there!

Elon: FSD Beta tweets

That’s actually a good definition. Should be 10x better than humans (so one minor accident like curbing once in 100,000 miles and serious accident once in a million miles). Only then would it be worthwhile for Tesla to take liability.

That’s why I’d be shocked if Tesla took liability for FSD anytime soon.

 

[jebinc]

That’s actually a good definition. Should be 10x better than humans (so one minor accident like curbing once in 100,000 miles and serious accident once in a million miles). Only then would it be worthwhile for Tesla to take liability.

That’s why I’d be shocked if Tesla took liability for FSD anytime soon.

I doubt they will ever take liability. L2 for the foreseeable future.

 

[FSDtester#1]

I doubt they will ever take liability. L2 for the foreseeable future.

Agreed, the stockholders and board members wouldn't want that amount of risk to the company even if it meant higher margins and higher fsd take rate. Some company in the future will take that leap, get sued into oblivion and go bankrupt. Let that company take the first punch in the face. I don't even want anyone to talk about the bs Benz level 3. It is stupid, and I trust fsd more in the situations it's allowed to operate. That tells you how bad MB is currently.

 

[Yelobird]

Agreed, the stockholders and board members wouldn't want that amount of risk to the company even if it meant higher margins and higher fsd take rate. Some company in the future will take that leap, get sued into oblivion and go bankrupt. Let that company take the first punch in the face. I don't even want anyone to talk about the bs Benz level 3. It is stupid, and I trust fsd more in the situations it's allowed to operate. That tells you how bad MB is currently.

My Guess is there is a reason Tesla unlike any other OEM opening an auto insurance company while scaling. IF that liability trigger is ever required I would bet it would only be covered using Their direct insurance. Insurance covers more then a direct accident so I suspect they can cover the random occurrence failure from the policy overage.

 

[FSDtester#1]

My Guess is there is a reason Tesla unlike any other OEM opening an auto insurance company while scaling. IF that liability trigger is ever required I would bet it would only be covered using Their direct insurance. Insurance covers more then a direct accident so I suspect they can cover the random occurrence failure from the policy overage.

Bingo... best answer !!

You nailed it. That is the long term plan. Great insight!!

Now if we could all get Tesla insurance in our states, that would be great!