Profound progress towards FSD

[cucubits]

But there's also radar in front of the Tesla. That should be able to distinguish between a stopped object not in the travel lane from a stopped object in the travel lane. If it can't then a higher definition radar should be installed. This is a "can't do without" issue. Another issue that a self-driving car can't do without is the ability to see deer and other animals in or near the travel lane. A human, seeing a deer on the side of the road, or crossing the road, knows there may be other deer nearby and slows down. FSD needs to do the same. It has to realize that a deer standing perfectly still might suddenly jump in front of the car. Or one of the members of its family, which might not be visible, might suddenly jump in front of the car.

Another FSD issue is potholes. Right now, I understand that FSD can't distinguish between an actual pothole and a patched pothole so it ignores potholes. FSD is going to have to be able to make that distinction.

All of these are valid observations and FSD is nowhere near having any of these solved, especially with the current limited sensor set it uses.

It's been said many times... Elon is a great salesman. He's managed to convince lots of folks to believe in FSD and pay $8k for it now, without any hard evidence that it will do what's being promised for quite some years now.

 

[diplomat33]

But there's also radar in front of the Tesla. That should be able to distinguish between a stopped object not in the travel lane from a stopped object in the travel lane. If it can't then a higher definition radar should be installed. This is a "can't do without" issue. Another issue that a self-driving car can't do without is the ability to see deer and other animals in or near the travel lane. A human, seeing a deer on the side of the road, or crossing the road, knows there may be other deer nearby and slows down. FSD needs to do the same. It has to realize that a deer standing perfectly still might suddenly jump in front of the car. Or one of the members of its family, which might not be visible, might suddenly jump in front of the car.

Another FSD issue is potholes. Right now, I understand that FSD can't distinguish between an actual pothole and a patched pothole so it ignores potholes. FSD is going to have to be able to make that distinction.

I don't think radar has enough resolution, at least not the ones that Tesla uses. So radar can't identify a deer or a pothole. Radar will give you a lot of returns from static and moving objects. You to have ignore all the static objects because otherwise your car would phantom brake a lot. But eliminating all static objects, means that your car won't "see" the stopped cars in your lane or next to your lane. Radar is great for adaptive cruise control because once you eliminate all the static objects, you are just left with moving objects that you do care about. And the radar will tell you the precise distance and speed of the car in front of you, allowing you to maintain a safe distance from the car.

Lidar is excellent for these cases. lidar will detect all objects, including static objects. and lidar has the resolution to actually identify a deer or the exact size and position of a pothole. Lidar can even detect the movement of the deer's head or the slight movement of its legs that you can use to estimate that the deer might jump in front of you.

 

[Cheburashka]

But there's also radar in front of the Tesla. That should be able to distinguish between a stopped object not in the travel lane from a stopped object in the travel lane. If it can't then a higher definition radar should be installed.

Radar doesn't really work that way. This is why cars have LIDAR.

Edit: diplomat33 beat me to it with a better explanation

 

[HenryFarkas]

I don't think radar has enough resolution, at least not the ones that Tesla uses. So radar can't identify a deer or a pothole. Radar will give you a lot of returns from static and moving objects. You to have ignore all the static objects because otherwise your car would phantom brake a lot. But eliminating all static objects, means that your car won't "see" the stopped cars in your lane or next to your lane. Radar is great for adaptive cruise control because once you eliminate all the static objects, you are just left with moving objects that you do care about. And the radar will tell you the precise distance and speed of the car in front of you, allowing you to maintain a safe distance from the car.

Lidar is excellent for these cases. lidar will detect all objects, including static objects. and lidar has the resolution to actually identify a deer or the exact size and position of a pothole. Lidar can even detect the movement of the deer's head or the slight movement of its legs that you can use to estimate that the deer might jump in front of you.

So basically, Elon Musk was lying to us when he said that his neural network would lead to FSD, and that we'd be able to send our cars out to work as taxis when we didn't need them ourselves.

 

[cucubits]

So basically, Elon Musk was lying to us when he said that his neural network would lead to FSD, and that we'd be able to send our cars out to work as taxis when we didn't need them ourselves.

Yes and no. This probably will be true sometimes in the next 10-20 years.

 

[HenryFarkas]

Yes and no. This probably will be true sometimes in the next 10-20 years.

But not without an effective sensor suite. Apparently, that means lidar, and Musk has repeatedly claimed that lidar is too expensive and is not needed for FSD. That's the lie. His claim is that he's going to achieve FSD without lidar.

 

[Mardak]

No it doesn't mean palo alto road were safer. Demo was strictly demo.

I think we're saying the same thing. I wasn't suggesting Palo Alto roads were generally safe enough to publicly deploy a FSD update to owners -- just that Tesla knew their FSD Autopilot would work good enough for the roads likely to be demoed to passengers.

So it seems like Autopilot team has some level of "safe enough" overall for a country to avoid releasing features that are too often more dangerous in certain regions. But even then, the driver would need to actively monitor and take over, so I wonder if the threshold is at least say… 50% for a given feature. E.g., does the car correctly detect "your" red light at least 50% of the time to turn on running-red-light warning or perhaps that threshold is lower for a warning and higher for actual driving changes?

 

[ItsNotAboutTheMoney]

But not without an effective sensor suite. Apparently, that means lidar, and Musk has repeatedly claimed that lidar is too expensive and is not needed for FSD. That's the lie. His claim is that he's going to achieve FSD without lidar.

If he believes that vision is sufficient, it's not a lie.

 

[diplomat33]

So basically, Elon Musk was lying to us when he said that his neural network would lead to FSD, and that we'd be able to send our cars out to work as taxis when we didn't need them ourselves.

The industry consensus is that lidar is required because no one sensor can do it all:

"As of today, a single sensor is not capable of simultaneously providing reliable and precise detection, classifications, measurements, and robustness to adverse conditions. Therefore, a multimodal approach is required to cover the detectability of relevant entities. "

Source: Safety First for Automated Driving. page 47.

But Elon is absolutely convinced that cameras and neural networks are enough to solve autonomous driving. So much so that he acts like it is already a done deal. He promises and sells FSD, declares FSD a "solved problem", declares unequivocally that "lidar is doomed!" simply because he believes in his approach, even though he has not achieved FSD yet.

It is one thing to believe in your approach and to try a different approach. What I think is deceptive though is that he is promising and selling a future product that he does not have yet. He is selling FSD to customers even though Tesla does not have FSD yet.

 

[diplomat33]

If he believes that vision is sufficient, it's not a lie.

 

[Buckminster]

Lidar is not needed. FSD is a solved problem - merely need to walk through the use cases and the 9s.

Teslas are already 10 times safer than other cars.

 

[diplomat33]

Lidar is not needed. FSD is a solved problem - merely need to walk through the use cases and the 9s.

Teslas are already 10 times safer than other cars.

How is FSD a solved problem? Tesla has not solved FSD.

You are like Elon. You are acting like FSD is a done deal. It's not. Tesla has not solved FSD and Tesla has only scratched the surface of edge cases that need to be solved.

And yes, Teslas are 10x safer but that is factoring in both Autopilot AND the human driver working together. Tesla's FSD with no human is not 10x safer. I guarantee you that if you put a Tesla with no human driver in it, it would not be 10x safer.

 

[Buckminster]

How is FSD a solved problem? Tesla has not solved FSD.

You are like Elon. You are acting like FSD is a done deal. It's not. Tesla has not solved FSD and Tesla has only scratched the surface of edge cases that need to be solved.

And yes, Teslas are 10x safer but that is factoring in both Autopilot AND the human driver working together. Tesla's FSD with no human is not 10x safer. I guarantee you that if you put a Tesla with no human driver in it, it would not be 10x safer.

Which edge case worries you?

 

[diplomat33]

Which edge case worries you?

There are too many to mention. There are thousands of edge cases that Tesla has not solved yet.

We know that Tesla has not solved obvious problems like a white truck pulling in front of you or a stopped vehicle on the side of the road.

But probably what worry me the most are cases where cameras are likely to fail.

For examples:
- Cameras being blinded
- Cameras not seeing quick enough a pedestrian with dark clothes or dark animal at night
- Cameras not seeing quick enough a pedestrian hiding in a shadow like my dark tunnel example.
- Cameras not recognizing sharp glass on the road because glass is see-through.

 

[HenryFarkas]

It's certainly true that without a human driver, the Tesla would crash nearly every time. At least, that's been my experience in the month and a half that I've owned my M3. And that's during a pandemic when I'm not driving much, and there's not much traffic. It's the exception rather than the rule that I can do an entire drive on NoA without having to intervene to prevent an accident. I can't figure out how people got enthusiastic about FSD in the past when, presumably FSD was not as capable as it is now.

 

[mikes_fsd]

It's certainly true that without a human driver, the Tesla would crash nearly every time. At least, that's been my experience in the month and a half that I've owned my M3. And that's during a pandemic when I'm not driving much, and there's not much traffic. It's the exception rather than the rule that I can do an entire drive on NoA without having to intervene to prevent an accident. I can't figure out how people got enthusiastic about FSD in the past when, presumably FSD was not as capable as it is now.

Clearly, you buy things without reading what you're buying. Or maybe reading and selectively deciding what you're going to believe.
The "what" and the "how long" are listed, but if you read these forums you'll find this little block of text, sliced and diced to fit their narrative.

What:

• "full self-driving in almost all circumstances"

How long:

• "future use of these features without supervision is dependent on achieving reliability far in excess of human drivers..."

 

[Buckminster]

For examples:
- Cameras being blinded
- Cameras not seeing quick enough a pedestrian with dark clothes or dark animal at night
- Cameras not seeing quick enough a pedestrian hiding in a shadow like my dark tunnel example.
- Cameras not recognizing sharp glass on the road because glass is see-through.

Camera and computer combo should be no worse than a human who is paying attention 100% for each of these. Humans do not pay attention 100%. Radar makes FSD better than humans for all of these except the glass.

 

[Cheburashka]

Camera and computer combo should be no worse than a human who is paying attention 100% for each of these. Humans do not pay attention 100%. Radar makes FSD better than humans for all of these except the glass.

Not really since it can't see stationary objects or road debris.

 

[diplomat33]

Camera and computer combo should be no worse than a human who is paying attention 100% for each of these. Humans do not pay attention 100%. Radar makes FSD better than humans for all of these except the glass.

The problem is that "camera and computer combo" is not as good as "human vision and brain combo". They are not equal. And radar would not help in those cases I mentioned. If the pedestrian were stationary, the radar would probably ignore it.

 

[Buckminster]

The problem is that "camera and computer combo" is not as good as "human vision and brain combo". They are not equal. And radar would not help in those cases I mentioned. If the pedestrian were stationary, the radar would probably ignore it.

Stationary pedestrian in a tunnel that jumps in front of the car = TSLAQ