6 ms·

What does Waymo lack in your opinion to not be considered "full self driving"? The persistent problem seems to be severe weather, but the gap between the weath

What does Waymo lack in your opinion to not be considered "full self driving"? The persistent problem seems to be severe weather, but the gap between the weather a human shouldn't drive in and weather a robot can't drive in will only get smaller. In the end, the reason to own a self-driven vehicle may come down to how many severe weather days you have to endure in your locale.

panick21_ 1y ago
Put a Waymo on random road in the world, can it drive it?
- cryptoz 1y ago
  Many humans couldn't.
  - jacquesm 1y ago
    Most humans that claim they could could. Anyway, this seems like a pretty low quality comment, you got perfectly well what the OP meant.
    - cryptoz 1y ago
      Oh gosh sorry, I do try to contribute positively to HN and write quality comments. I'll expand: I've been in circumstances where I've been rented a company car in a foreign country, felt that I was a good driver, but struggled. The road signs are different and can be confusing, the local patterns and habits of drivers can be totally different from what you're accustomed to. I don't doubt that lots of humans could drive most roads - but I think the average driver would struggle, and have a much higher rate of accidents than a local. Germany, Italy, India all stand out as examples to me. The roads and driving culture is very different, and can be dangerous to someone who is used to driving on American suburban streets. I really do stand by my comment, and apologize for the 'low quality' nature of it. I meant to suggest that we set the bar far higher for AI than we do for people, which is in general a good thing. But still - I would say that by this definition of 'full self driving', it wouldn't be met very well by many or most human drivers.
      - jacquesm 1y ago
        I've driven all over the planet except for Asia and Africa. So far, no real problem and I think most drivers would adapt within a day or two. Greece, Panama and Colombia stand out as somewhat more exciting. Switching to left hand driving in the UK also wasn't a big problem but you do have to pay more attention. Of course I may have simply been lucky, but given that my driving license is valid in many countries it seems as though humanity has determined this is mostly a solved problem. When someone says "Put a Waymo on random road in the world, can it drive it?" they mean: I would expect a human to be able to drive on a random road in the world. And they likely could. Can a Waymo do the same? I don't know the answer to that one. But if there is one thing that humans are pretty good at it is adaptation to circumstances previously unseen. I am not sure if a Waymo could do the same but it would be a very interesting experiment to find out. American suburban streets are not representative of driving in most parts of the world. I don't think the bar of 'should be able to drive most places where humans can drive' is all that high and even your average American would adapt pretty quickly to driving in different places. Source: I know plenty of Americans and have seen them drive in lots of countries. Usually it works quite well, though, admittedly, seeing them in Germany was kind of funny. "Am I hallucinating or did we just get passed by an old lady? And we're doing 85 Mph?"
      - gerdesj 1y ago
        "Germany, Italy, India " That's experience and you learned and survived to tell the tale. Its almost as though you are capable of learning how to deal with an unfamiliar environment, and fail safe! I'm a Brit and have driven across most of Europe, US/CA and a few other places. Southern Italy eg around Napoli is pretty fraught - around there I find that you need to treat your entire car as an indicator: if you can wedge your car into a traffic stream, you will be let in, mostly without horns blaring. If you sit and wait, you will go grey haired eventually. In Germania, speed is king. I lived there in the 70s-90s as well as being a visitor recently. The autobahns are insane if you stray out of lane one, the rest of the road system is civilised. France - mostly like driving around the UK apart from their weird right hand side of the road thing! La Perifique is just as funky as the M25 and La Place du Concorde is a right old laugh. The rest of the country that I have driven is very civilised. Europe to the right of Italy is pretty safe too. I have to say that across the entirety of Europe, that road signage is very good. The one sign that might confuse any non-European is the white and yellow diamond (we don't have them in the UK). It means that you have priority over an implied "priority to the right". See https://driveeurope.co.uk/2013/02/27/priority-to-the-right/ https://driveeurope.co.uk/2013/02/27/priority-to-the-right/ for a decent explanation. Roundabouts were invented in the US. In the UK when you are actually on a roundabout you have right of way. However, everyone will behave as though "priorite a la doite" and there will often be a stand off - its hilarious! In the UK, when someone flashes their headlights at you it generally means "I have seen you and will let you in". That generally surprises foreigners (I once gave a lift to a prospective employee candidate from Poland and he was absolutely aghast at how polite our roads seemed to be). Don't always assume that you will be given space but we are pretty good at "after you".
    - bsder 1y ago
      > Most humans that claim they could could. I don't agree. My anecdata suggests that Waymo is significantly better than random ridesharing drivers in the US, nowadays. My last dozen ridesharing experiences only had a single driver that wasn't actively hazardous on the road. One of them was so bad that I actually flagged him on the service. My Waymo experiences, by contrast, have all been uniformly excellent. I suspect that Waymo is already better than the median human driver (anecdata suggests that's a really low bar)--and it just keeps getting better.
      - jacquesm 1y ago
        > Most humans that claim they could could. > My anecdata suggests that Waymo is significantly better than random ridesharing drivers in the US, nowadays. Those two aren't really related are they? That's one locality and a specific kind of driver. If you picked a random road there is a pretty small chance that road would be one like the one where Waymo is currently rolled out, and where your ridesharing drivers are representative of the general public, they likely are not.
- Kye 1y ago
  That's the real issue. If "can navigate roads" is enough then we've had full self-driving for a while. There needs to be some base level of general purpose capability or it's just a neat regional curiosity.
- standardUser 1y ago
  For a couple decades you couldn't even bring your cell phone anywhere in the world and use it. Transformational technologies don't have to be available universally and simultaneously to be viable. Even when the gas car was created you couldn't use it anywhere that didn't have gasoline and paved roads, plus a mechanic and access to parts.
  - jazzyjackson 1y ago
    A significant portion of US highways and backroads are uncovered by cell signal. I suppose a self driving car would have starlink these days.
    - standardUser 1y ago
      We once had no gas stations, now we have 150,000 (in the US). If the commercial need is there, building out connectivity is an unlikely impediment. Starlink et al. can solve this everywhere except when there's severe weather, a problem Waymo shares, which is starting to make me think the Upper Midwest might be waiting a very long time for self-driving cars.
      - panick21_ 1y ago
        I think the bigger problem is mapping every road to the detail they need and keeping that up to date.
  - panick21_ 1y ago
    Did I argue that the technology was not viable? I answered the question 'What does Waymo lack in your opinion to not be considered "full self driving"?'. And clearly its not if it can't drive on literally 99.99% of roads in the world. Any argument to the contrary is just ridiculous.
    - standardUser 1y ago
      As I said above, "full self driving" is clearly an outdated concept.
mkl 1y ago
Waymo is very restricted on the locations it drives (limit parts of limited cities, I think no freeways still), and uses remote operators to make decisions in unusual situations and when it gets stuck. This article from last year has quite a bit of information: https://arstechnica.com/cars/2024/05/on-self-driving-waymo-is-playing-chess-while-tesla-plays-checkers/ https://arstechnica.com/cars/2024/05/on-self-driving-waymo-i...
- panarky 1y ago
  Waymo never allows a remote human to drive the car. If it gets stuck, a remote operator can assess the situation and tell the car where it should go, but all driving is always handled locally by the onboard system in the vehicle. Interesting that Waymo now operates just fine in SF fog, and is expanding to Seattle (rain) and Denver (snow and ice).
  - epcoa 1y ago
    The person you're replying to never claimed otherwise. However, while decision support is not directly steering and accelerating/braking the car, I am just going to assert it is still driving the car, at least for how it actually matters in this discussion. And the best estimate is that these interventions are "uncommon" on the order of 10ks miles, but that isn't rare. A system that requires a "higher level" handler is not full self driving.
    - panarky 1y ago
      Agreed! Which is why an autonomous car company that is responsible and prioritizes safety would never call their SAE Level 4 vehicle "full self-driving". And that's why it's so irresponsible and dangerous for Tesla to continue using that marketing hype term for their SAE Level 2 system.
    - AlotOfReading 1y ago
      There's a simple test I find useful to determine who's driving: If the vehicle has a collision, who's ultimately responsible? That person (or computer) is the driver. If a Waymo hits a pole for example, the software has a bug. It wasn't the responsibility of a remote assistant to monitor the environment in real time and prevent the accident, so we call the computer the driver. If we put a safety driver in the seat and run the same software that hits the same pole, it was the human who didn't meet their responsibility to prevent the accident. Therefore, they're the driver.
gerdesj 1y ago
No one does FSD yet - properly. It initially seems mad that a human, inside the box can outperform the "finest" efforts of a multi zillion dollar company. The human has all their sensors inside the box and most of them stymied by the non transparent parts. Bad weather makes it worse. However, look at the sensors and compute being deployed on cars. Its all minimums and cost focused - basically MVP, with deaths as a costed variable in an equation. A car could have cameras with views everywhere for optical, LIDAR, RADAR, even a form of SONAR if it can be useful, microwave and way more. Accellerometers and all sorts too, all feeding into a model. As a driver, I've come up with strategies such as "look left, listen right". I'm British so drive on the left and sit on the right side of my car. When turning right and I have the window wound down, I can watch the left for a gap and listen for cars to the right. I use it as a negative and never a positive - so if I see a gap on the left and I hear a car to my right, I stay put. If I see a gap to the left but hear no sound on my right, I turn my head to confirm that there is a space and do a final quick go/no go (which involves another check left and right). This strategy saves quite a lot of head swings and if done properly is safe. I now drive an EV: One year so far - a Seic MG4, with cameras on all four sides, that I can't record from but can use. It has lane assist (so lateral control, which craps out on many A road sections but is fine on motorway class roads) and cruise control that will keep a safe distance from other vehicles (that works well on most roads and very well on motorways, there are restrictions). Recently I was driving and a really heavy rain shower hit as I was overtaking a lorry. I immediately dived back into lane one, behind the lorry and put cruise on. I could just see the edge white line, so I dealt with left/right and the car sorted out forward/backward. I can easily deal with both but its quite nice to be able carefully abrogate responsibilities.

panick21_ 1y ago

Put a Waymo on random road in the world, can it drive it?

cryptoz 1y ago
Many humans couldn't.
- jacquesm 1y ago
  Most humans that claim they could could. Anyway, this seems like a pretty low quality comment, you got perfectly well what the OP meant.
  - cryptoz 1y ago
    Oh gosh sorry, I do try to contribute positively to HN and write quality comments. I'll expand: I've been in circumstances where I've been rented a company car in a foreign country, felt that I was a good driver, but struggled. The road signs are different and can be confusing, the local patterns and habits of drivers can be totally different from what you're accustomed to. I don't doubt that lots of humans could drive most roads - but I think the average driver would struggle, and have a much higher rate of accidents than a local. Germany, Italy, India all stand out as examples to me. The roads and driving culture is very different, and can be dangerous to someone who is used to driving on American suburban streets. I really do stand by my comment, and apologize for the 'low quality' nature of it. I meant to suggest that we set the bar far higher for AI than we do for people, which is in general a good thing. But still - I would say that by this definition of 'full self driving', it wouldn't be met very well by many or most human drivers.
    - jacquesm 1y ago
      I've driven all over the planet except for Asia and Africa. So far, no real problem and I think most drivers would adapt within a day or two. Greece, Panama and Colombia stand out as somewhat more exciting. Switching to left hand driving in the UK also wasn't a big problem but you do have to pay more attention. Of course I may have simply been lucky, but given that my driving license is valid in many countries it seems as though humanity has determined this is mostly a solved problem. When someone says "Put a Waymo on random road in the world, can it drive it?" they mean: I would expect a human to be able to drive on a random road in the world. And they likely could. Can a Waymo do the same? I don't know the answer to that one. But if there is one thing that humans are pretty good at it is adaptation to circumstances previously unseen. I am not sure if a Waymo could do the same but it would be a very interesting experiment to find out. American suburban streets are not representative of driving in most parts of the world. I don't think the bar of 'should be able to drive most places where humans can drive' is all that high and even your average American would adapt pretty quickly to driving in different places. Source: I know plenty of Americans and have seen them drive in lots of countries. Usually it works quite well, though, admittedly, seeing them in Germany was kind of funny. "Am I hallucinating or did we just get passed by an old lady? And we're doing 85 Mph?"
    - gerdesj 1y ago
      "Germany, Italy, India " That's experience and you learned and survived to tell the tale. Its almost as though you are capable of learning how to deal with an unfamiliar environment, and fail safe! I'm a Brit and have driven across most of Europe, US/CA and a few other places. Southern Italy eg around Napoli is pretty fraught - around there I find that you need to treat your entire car as an indicator: if you can wedge your car into a traffic stream, you will be let in, mostly without horns blaring. If you sit and wait, you will go grey haired eventually. In Germania, speed is king. I lived there in the 70s-90s as well as being a visitor recently. The autobahns are insane if you stray out of lane one, the rest of the road system is civilised. France - mostly like driving around the UK apart from their weird right hand side of the road thing! La Perifique is just as funky as the M25 and La Place du Concorde is a right old laugh. The rest of the country that I have driven is very civilised. Europe to the right of Italy is pretty safe too. I have to say that across the entirety of Europe, that road signage is very good. The one sign that might confuse any non-European is the white and yellow diamond (we don't have them in the UK). It means that you have priority over an implied "priority to the right". See https://driveeurope.co.uk/2013/02/27/priority-to-the-right/ https://driveeurope.co.uk/2013/02/27/priority-to-the-right/ for a decent explanation. Roundabouts were invented in the US. In the UK when you are actually on a roundabout you have right of way. However, everyone will behave as though "priorite a la doite" and there will often be a stand off - its hilarious! In the UK, when someone flashes their headlights at you it generally means "I have seen you and will let you in". That generally surprises foreigners (I once gave a lift to a prospective employee candidate from Poland and he was absolutely aghast at how polite our roads seemed to be). Don't always assume that you will be given space but we are pretty good at "after you".
  - bsder 1y ago
    > Most humans that claim they could could. I don't agree. My anecdata suggests that Waymo is significantly better than random ridesharing drivers in the US, nowadays. My last dozen ridesharing experiences only had a single driver that wasn't actively hazardous on the road. One of them was so bad that I actually flagged him on the service. My Waymo experiences, by contrast, have all been uniformly excellent. I suspect that Waymo is already better than the median human driver (anecdata suggests that's a really low bar)--and it just keeps getting better.
    - jacquesm 1y ago
      > Most humans that claim they could could. > My anecdata suggests that Waymo is significantly better than random ridesharing drivers in the US, nowadays. Those two aren't really related are they? That's one locality and a specific kind of driver. If you picked a random road there is a pretty small chance that road would be one like the one where Waymo is currently rolled out, and where your ridesharing drivers are representative of the general public, they likely are not.
Kye 1y ago
That's the real issue. If "can navigate roads" is enough then we've had full self-driving for a while. There needs to be some base level of general purpose capability or it's just a neat regional curiosity.
standardUser 1y ago
For a couple decades you couldn't even bring your cell phone anywhere in the world and use it. Transformational technologies don't have to be available universally and simultaneously to be viable. Even when the gas car was created you couldn't use it anywhere that didn't have gasoline and paved roads, plus a mechanic and access to parts.
- jazzyjackson 1y ago
  A significant portion of US highways and backroads are uncovered by cell signal. I suppose a self driving car would have starlink these days.
  - standardUser 1y ago
    We once had no gas stations, now we have 150,000 (in the US). If the commercial need is there, building out connectivity is an unlikely impediment. Starlink et al. can solve this everywhere except when there's severe weather, a problem Waymo shares, which is starting to make me think the Upper Midwest might be waiting a very long time for self-driving cars.
    - panick21_ 1y ago
      I think the bigger problem is mapping every road to the detail they need and keeping that up to date.
- panick21_ 1y ago
  Did I argue that the technology was not viable? I answered the question 'What does Waymo lack in your opinion to not be considered "full self driving"?'. And clearly its not if it can't drive on literally 99.99% of roads in the world. Any argument to the contrary is just ridiculous.
  - standardUser 1y ago
    As I said above, "full self driving" is clearly an outdated concept.

mkl 1y ago

Waymo is very restricted on the locations it drives (limit parts of limited cities, I think no freeways still), and uses remote operators to make decisions in unusual situations and when it gets stuck. This article from last year has quite a bit of information: https://arstechnica.com/cars/2024/05/on-self-driving-waymo-is-playing-chess-while-tesla-plays-checkers/ https://arstechnica.com/cars/2024/05/on-self-driving-waymo-i...

panarky 1y ago
Waymo never allows a remote human to drive the car. If it gets stuck, a remote operator can assess the situation and tell the car where it should go, but all driving is always handled locally by the onboard system in the vehicle. Interesting that Waymo now operates just fine in SF fog, and is expanding to Seattle (rain) and Denver (snow and ice).
- epcoa 1y ago
  The person you're replying to never claimed otherwise. However, while decision support is not directly steering and accelerating/braking the car, I am just going to assert it is still driving the car, at least for how it actually matters in this discussion. And the best estimate is that these interventions are "uncommon" on the order of 10ks miles, but that isn't rare. A system that requires a "higher level" handler is not full self driving.
  - panarky 1y ago
    Agreed! Which is why an autonomous car company that is responsible and prioritizes safety would never call their SAE Level 4 vehicle "full self-driving". And that's why it's so irresponsible and dangerous for Tesla to continue using that marketing hype term for their SAE Level 2 system.
  - AlotOfReading 1y ago
    There's a simple test I find useful to determine who's driving: If the vehicle has a collision, who's ultimately responsible? That person (or computer) is the driver. If a Waymo hits a pole for example, the software has a bug. It wasn't the responsibility of a remote assistant to monitor the environment in real time and prevent the accident, so we call the computer the driver. If we put a safety driver in the seat and run the same software that hits the same pole, it was the human who didn't meet their responsibility to prevent the accident. Therefore, they're the driver.

gerdesj 1y ago

No one does FSD yet - properly. It initially seems mad that a human, inside the box can outperform the "finest" efforts of a multi zillion dollar company. The human has all their sensors inside the box and most of them stymied by the non transparent parts. Bad weather makes it worse. However, look at the sensors and compute being deployed on cars. Its all minimums and cost focused - basically MVP, with deaths as a costed variable in an equation. A car could have cameras with views everywhere for optical, LIDAR, RADAR, even a form of SONAR if it can be useful, microwave and way more. Accellerometers and all sorts too, all feeding into a model. As a driver, I've come up with strategies such as "look left, listen right". I'm British so drive on the left and sit on the right side of my car. When turning right and I have the window wound down, I can watch the left for a gap and listen for cars to the right. I use it as a negative and never a positive - so if I see a gap on the left and I hear a car to my right, I stay put. If I see a gap to the left but hear no sound on my right, I turn my head to confirm that there is a space and do a final quick go/no go (which involves another check left and right). This strategy saves quite a lot of head swings and if done properly is safe. I now drive an EV: One year so far - a Seic MG4, with cameras on all four sides, that I can't record from but can use. It has lane assist (so lateral control, which craps out on many A road sections but is fine on motorway class roads) and cruise control that will keep a safe distance from other vehicles (that works well on most roads and very well on motorways, there are restrictions). Recently I was driving and a really heavy rain shower hit as I was overtaking a lorry. I immediately dived back into lane one, behind the lorry and put cruise on. I could just see the edge white line, so I dealt with left/right and the car sorted out forward/backward. I can easily deal with both but its quite nice to be able carefully abrogate responsibilities.