5 ms·

Comparison with human ability aside, this would most definitely not have happened with Lidar. Musk's argument has always been that you don't need Lidar because

Comparison with human ability aside, this would most definitely not have happened with Lidar. Musk's argument has always been that you don't need Lidar because humans don't need Lidar, but since the goal is to do better than humans this might not really work out in the end.

dkonofalski 5y ago
Agreed wholeheartedly. This is a perfect example of why computers should be assisted by things that humans aren't.
soheil 5y ago
Please try not to draw general conclusions from specific examples. First we need to know what percentage of crashes are due to lack of lidar/radar instead of looking at anecdotal cases.
- trompetenaccoun 5y ago
  Apart from the higher price why would you not want additional sensors like radar and lidar? If all human driven cars were already equipped with radar and emergency stop systems, mass pileups would be a thing of the past. Obviously there are situations where cameras or our eyes simply won't be able to see enough. Dark unlit objects at night aren't exactly an obscure edge case scenario, the crash from the video is something we already see on a regular basis with human driven cars, sadly. And even if not, I'd rather pay $1000 or so extra to protect me from that freak scenario that's likely never going to happen. Until it does. I also have airbags in my car although I never needed them.
  - soheil 5y ago
    If there are so many lidars near each other the chance of lasers causing constructive interference increases causing irreversible eye damage.
    - metadat 5y ago
      This appears to be unfounded speculation, do you have a source? Even if it is a real issue, the government regulators can simply set the max power limits lower for LiDAR units to ensure it's safe.
      - soheil 5y ago
        This is a well known and basic topic that you will be trained on before allowed to enter any physics lab with lasers in it. This can happen around shiny objects, windows on streets and pretty much anything that reflects light. "So coherent is laser light that one may observe interference effects even when the path difference between the interfering rays is much greater than 109 wavelengths! ... If two waves are exactly “in phase” as shown in Figure 1a, they will reinforce each other; this is called constructive interference." [0] [0] https://faraday.physics.utoronto.ca/IYearLab/intdif.pdf https://faraday.physics.utoronto.ca/IYearLab/intdif.pdf
      - trompetenaccoun 5y ago
        I was skeptical as well but looking it up it seems it's a genuine concern. Also you're right about limiting the power but interference in general could lead to problems, not just with lidar but radar as well. Maybe not impossible to solve but it seems like difficult issue, you wouldn't want misleading sensor reading lead to the car making a fatal decision.
Someone1234 5y ago
Even if it was only used for AEB instead of AutoPilot (i.e. acting as an actual backup to autopilot) that would be a vast improvement. A Tesla vehicle, when under human control, has AEB but when under AutoPilot it essentially does not since it is the same logic used for both processes and therefore it doesn't act as an additional safety feature.
nicoburns 5y ago
Yep, their bet on not using Lidar seems bizarre to me. Presumably it’s been done for cost purposes, but lidar will likely be cheap by the time AI driving is viable.
- ashtonbaker 5y ago
  I think the issue at this point is that they've promised - i.e. taken money for - "full self driving" features that are supposed to be delivered using the car's hardware as-is. So they have to either work with the cameras they have, or retrofit a bunch of cars with LIDAR, or just fail I guess.
soheil 5y ago
Machines are already faster/better in reaction times, not getting fatigued, etc. than humans why do you need more reasons to be better than humans?
- camjohnson26 5y ago
  Because of tradeoffs. A brick is more resilient than a human, that doesn’t make it better.
  - soheil 5y ago
    The items I listed above are attributes that make driving better whether you posses them or a brick.
davesque 5y ago
However, the car was equipped with radar. Twitter author also said that the truck was seen by radar. I don't see how something like this could have been missed by a radar-equipped car. I chalk it up to the black-box nature of these models and the general lack of awareness about what bizarre edge cases could be encoded in the model's parameters.
- fblp 5y ago
  How is black truck stuck in your lane an edge case? Seems more like negligence.
  - davesque 5y ago
    I think you're completely misunderstanding my comment. I was saying that the fact that the model missed the truck can be described as an edge case. It's an edge case in the sense that it could end up being a strange and rare outcome that results from an unfortunate compounding of error at the intersection of a few parameter in the model. Deep learning models (which are undoubtedly involved) are problematic in this way. Of course I agree that this sort of driving condition is not an edge case and should be in the training set. But that doesn't mean that anything like it will always be correctly categorized by the model. We've seen that it's possible to trick some computer vision models with well-placed, single pixels. If anything, I'm saying that we shouldn't be entrusting our roads to deep learning models over a certain level of complexity.
ModernMech 5y ago
It also shouldn't have happened with a RADAR. From the second tweet: the truck was seen by both the radar and also could be clearly observed by the narrow camera from quite a distance away (pictures are spaced 1 second apart) This is more about beta quality robotics software being shipped to actual customers and used in production. If you're someone who won't even use a programming language or library in production before it reaches v1.0, this is about a million times worse than that. By the way, as a roboticist this whole situation is completely baffling to me. We go through such great lengths in our research to make our robots safe. Safety is a huge concern to our community. How is it that this company can just ship beta quality software to customers for use in production amongst the rest of us? In industry, we paint our robots orange and put them behind impenetrable cages due to safety concerns. Entire facilities are designed around the robots to make their operation safe. Operators are trained to use them. People working around them are trained to work along side them and observe proper safety protocols. Proper insurance is carried in case of accidents. And when an accident does occur, the situation is taken apart, analyzed, and changes are made so they won't happen again. Not in the marketplace though, it seems. Apparently it's okay to just code whatever you want; put the code on whatever machine you want, even if it's "beta" quality; ship it to whomever you want who can pay for it; provide little to no training; lie about the capabilities ("Full self driving" minus the full, and self, and driving. "Autopilot" that isn't.); and when people inevitably and predictably end up dead, apparently it's okay to just keep on pushing this software out without a shred of self reflection or humility. Because you're saving the world I guess? I just don't understand this culture.
a-dub 5y ago
> Musk's argument has always been that you don't need Lidar because humans don't need Lidar, but since the goal is to do better than humans this might not really work out in the end. aren't auto manufacturers that don't have features like autopilot currently adding things like lidar and automatic breaking to improve safety? why wouldn't you add the thing that will make your system as safe as possible? what's the goal here? some aspirational tech challenge that competes on arbitrary merits in some humans vs. machines challenge problem or building the best and safest way to move vehicles around?
- ModernMech 5y ago
  The issue is that the guy who runs the company and controls everything doesn't know what he doesn't know. He made the uninformed decision years ago to go all-in on camera technology because he thought that he would be able to throw a bunch of money and engineers at the problem, and solve Full Self Driving in essentially 5 years. That was more than 5 years ago and the technology seems to have plateaued at a place where it runs frequently into parked or stationary objects. They have since taken untold sums of cash from people (millions? tens of millions?) who they promised would get access to "Full Self Driving" with the hardware on board. Now they're in the situation where they cannot admit that the problem requires more hardware. The hardware is a fixed property here, a design requirement; they have to achieve actual full self driving with the hardware in place, or they're going to face refunds and lawsuits. So now it's a delay game. I think it's clear now that short of AGI, Tesla will never get to "full" self driving with the sensor suit they have. Nonetheless they can't switch to a new sensor suit because that will admit defeat. Instead, they have to show continual improvement of the current sensor suite, while promising that "Full" self driving is just around the corner so they can continue the charade.
  - avereveard 5y ago
    > they have to achieve actual full self driving with the hardware in place, or they're going to face refunds and lawsuits. they since retrofitted the hardware twice since the claim was made, last retrofit being to the tune of 10k$.
avereveard 5y ago
so is tesla running cameras with 21+ stops range? idk why isn't this argument dismantled every time musk mentions it just by pointing at how actually good eyes are.

dkonofalski 5y ago

Agreed wholeheartedly. This is a perfect example of why computers should be assisted by things that humans aren't.

soheil 5y ago

Please try not to draw general conclusions from specific examples. First we need to know what percentage of crashes are due to lack of lidar/radar instead of looking at anecdotal cases.

trompetenaccoun 5y ago
Apart from the higher price why would you not want additional sensors like radar and lidar? If all human driven cars were already equipped with radar and emergency stop systems, mass pileups would be a thing of the past. Obviously there are situations where cameras or our eyes simply won't be able to see enough. Dark unlit objects at night aren't exactly an obscure edge case scenario, the crash from the video is something we already see on a regular basis with human driven cars, sadly. And even if not, I'd rather pay $1000 or so extra to protect me from that freak scenario that's likely never going to happen. Until it does. I also have airbags in my car although I never needed them.
- soheil 5y ago
  If there are so many lidars near each other the chance of lasers causing constructive interference increases causing irreversible eye damage.
  - metadat 5y ago
    This appears to be unfounded speculation, do you have a source? Even if it is a real issue, the government regulators can simply set the max power limits lower for LiDAR units to ensure it's safe.
    - soheil 5y ago
      This is a well known and basic topic that you will be trained on before allowed to enter any physics lab with lasers in it. This can happen around shiny objects, windows on streets and pretty much anything that reflects light. "So coherent is laser light that one may observe interference effects even when the path difference between the interfering rays is much greater than 109 wavelengths! ... If two waves are exactly “in phase” as shown in Figure 1a, they will reinforce each other; this is called constructive interference." [0] [0] https://faraday.physics.utoronto.ca/IYearLab/intdif.pdf https://faraday.physics.utoronto.ca/IYearLab/intdif.pdf
    - trompetenaccoun 5y ago
      I was skeptical as well but looking it up it seems it's a genuine concern. Also you're right about limiting the power but interference in general could lead to problems, not just with lidar but radar as well. Maybe not impossible to solve but it seems like difficult issue, you wouldn't want misleading sensor reading lead to the car making a fatal decision.

Someone1234 5y ago

Even if it was only used for AEB instead of AutoPilot (i.e. acting as an actual backup to autopilot) that would be a vast improvement. A Tesla vehicle, when under human control, has AEB but when under AutoPilot it essentially does not since it is the same logic used for both processes and therefore it doesn't act as an additional safety feature.

nicoburns 5y ago

Yep, their bet on not using Lidar seems bizarre to me. Presumably it’s been done for cost purposes, but lidar will likely be cheap by the time AI driving is viable.

ashtonbaker 5y ago
I think the issue at this point is that they've promised - i.e. taken money for - "full self driving" features that are supposed to be delivered using the car's hardware as-is. So they have to either work with the cameras they have, or retrofit a bunch of cars with LIDAR, or just fail I guess.

soheil 5y ago

Machines are already faster/better in reaction times, not getting fatigued, etc. than humans why do you need more reasons to be better than humans?

camjohnson26 5y ago
Because of tradeoffs. A brick is more resilient than a human, that doesn’t make it better.
- soheil 5y ago
  The items I listed above are attributes that make driving better whether you posses them or a brick.

davesque 5y ago

However, the car was equipped with radar. Twitter author also said that the truck was seen by radar. I don't see how something like this could have been missed by a radar-equipped car. I chalk it up to the black-box nature of these models and the general lack of awareness about what bizarre edge cases could be encoded in the model's parameters.

fblp 5y ago
How is black truck stuck in your lane an edge case? Seems more like negligence.
- davesque 5y ago
  I think you're completely misunderstanding my comment. I was saying that the fact that the model missed the truck can be described as an edge case. It's an edge case in the sense that it could end up being a strange and rare outcome that results from an unfortunate compounding of error at the intersection of a few parameter in the model. Deep learning models (which are undoubtedly involved) are problematic in this way. Of course I agree that this sort of driving condition is not an edge case and should be in the training set. But that doesn't mean that anything like it will always be correctly categorized by the model. We've seen that it's possible to trick some computer vision models with well-placed, single pixels. If anything, I'm saying that we shouldn't be entrusting our roads to deep learning models over a certain level of complexity.

ModernMech 5y ago

It also shouldn't have happened with a RADAR. From the second tweet: the truck was seen by both the radar and also could be clearly observed by the narrow camera from quite a distance away (pictures are spaced 1 second apart) This is more about beta quality robotics software being shipped to actual customers and used in production. If you're someone who won't even use a programming language or library in production before it reaches v1.0, this is about a million times worse than that. By the way, as a roboticist this whole situation is completely baffling to me. We go through such great lengths in our research to make our robots safe. Safety is a huge concern to our community. How is it that this company can just ship beta quality software to customers for use in production amongst the rest of us? In industry, we paint our robots orange and put them behind impenetrable cages due to safety concerns. Entire facilities are designed around the robots to make their operation safe. Operators are trained to use them. People working around them are trained to work along side them and observe proper safety protocols. Proper insurance is carried in case of accidents. And when an accident does occur, the situation is taken apart, analyzed, and changes are made so they won't happen again. Not in the marketplace though, it seems. Apparently it's okay to just code whatever you want; put the code on whatever machine you want, even if it's "beta" quality; ship it to whomever you want who can pay for it; provide little to no training; lie about the capabilities ("Full self driving" minus the full, and self, and driving. "Autopilot" that isn't.); and when people inevitably and predictably end up dead, apparently it's okay to just keep on pushing this software out without a shred of self reflection or humility. Because you're saving the world I guess? I just don't understand this culture.

a-dub 5y ago

> Musk's argument has always been that you don't need Lidar because humans don't need Lidar, but since the goal is to do better than humans this might not really work out in the end. aren't auto manufacturers that don't have features like autopilot currently adding things like lidar and automatic breaking to improve safety? why wouldn't you add the thing that will make your system as safe as possible? what's the goal here? some aspirational tech challenge that competes on arbitrary merits in some humans vs. machines challenge problem or building the best and safest way to move vehicles around?

ModernMech 5y ago
The issue is that the guy who runs the company and controls everything doesn't know what he doesn't know. He made the uninformed decision years ago to go all-in on camera technology because he thought that he would be able to throw a bunch of money and engineers at the problem, and solve Full Self Driving in essentially 5 years. That was more than 5 years ago and the technology seems to have plateaued at a place where it runs frequently into parked or stationary objects. They have since taken untold sums of cash from people (millions? tens of millions?) who they promised would get access to "Full Self Driving" with the hardware on board. Now they're in the situation where they cannot admit that the problem requires more hardware. The hardware is a fixed property here, a design requirement; they have to achieve actual full self driving with the hardware in place, or they're going to face refunds and lawsuits. So now it's a delay game. I think it's clear now that short of AGI, Tesla will never get to "full" self driving with the sensor suit they have. Nonetheless they can't switch to a new sensor suit because that will admit defeat. Instead, they have to show continual improvement of the current sensor suite, while promising that "Full" self driving is just around the corner so they can continue the charade.
- avereveard 5y ago
  > they have to achieve actual full self driving with the hardware in place, or they're going to face refunds and lawsuits. they since retrofitted the hardware twice since the claim was made, last retrofit being to the tune of 10k$.

avereveard 5y ago

so is tesla running cameras with 21+ stops range? idk why isn't this argument dismantled every time musk mentions it just by pointing at how actually good eyes are.