4 ms·

Driverless cars are significantly safer than human drivers. https://arxiv.org/ftp/arxiv/papers/2309/2309.01206.pdf https://arxiv.org/ftp/arxiv/papers/2309/2309

by adrr 3y ago

Driverless cars are significantly safer than human drivers. https://arxiv.org/ftp/arxiv/papers/2309/2309.01206.pdf https://arxiv.org/ftp/arxiv/papers/2309/2309.01206.pdf

defrost 3y ago
"So far" With less than 4 million miles of pure rider only data by relatively early Waymo rollouts in two cities it can be argued there's room for the claims per million miles to climb as the number of deployments rise.
- kilobite88 3y ago
  Does it not drive the same roads as human drivers? 4 million miles = 4 million miles.
  - comte7092 3y ago
    > Does it not drive the same roads as human drivers? No, human drivers drive on roads beyond selected neighborhoods of Phoenix and San Francisco. It is unknown how generalizable the results of the study would be beyond the conditions experienced on the roads in the areas Waymo had been deployed in. We might see significantly improved safety or we might not. But given the messaging from Waymo, the presumption would seem to be that performance would be worse in other areas/conditions.
    - adrr 3y ago
      Are there more accidents on these roads?
      - comte7092 3y ago
        I am not really sure what your question is asking. More accidents on what roads? If you’re talking about the 99% of roads in the country where Waymo doesn’t currently operate, then I can’t really take your question seriously. We are talking about relative performance of an automated system vs human drivers. The automated system has been specifically tuned to perform well on a certain set of roads. The point is that there is no reason to assume by default that the system would maintain good relative performance under a drastically different set of conditions.
  - defrost 3y ago
    4 million miles is proportionally less than the width of a gnat against the history of driver miles in the US across all cities, less than a rounding error. More crucially, these are 4 million entry level service rollout miles .. a time when the company has highly paid engineers paying scrupulous detailed attention to the condition of the cars and the tuning of the automated responses, with jobs dedicated to the avoidance of PR incidents. comte7092 has pointed out other factors - but these are not "typical miles" these are the miles before typical sets in - who can say if the stats will be maintained as sensors age and maintainence becomes an after tought in the manner typical of human driven cars. I hope the figure holds up, I'm commenting here on the apples to oranges issues in the comparison at present.
comte7092 3y ago
I never said that they weren’t/wouldn’t be. But that wasn’t the point I was responding to. The point I was responding to was that we need to address 40,000 traffic deaths per year, to which I said it doesn’t seem to be the most direct route there. Just because self driving cars are safer in some situations doesn’t mean that we will inevitably have that result within a reasonable timeframe and level of investment. Besides the types of things that reduce human caused accidents would lead to benefits for self driving cars: slower speeds, fewer conflicts, etc.
- greiskul 3y ago
  > a reasonable timeframe and level of investment Compared to the time-frame and investments on which alternatives? It might take 40 years for self driving cars to improve and have enough deployments to significantly reduce the number of accidents, but at least there is a path there and it will happen at some point. Getting americans to use public transit, requires not only investment in public transit: it requires a cultural shift in how Americans view life. A cheaper solution that can never be delivered is not a good solution, it's just a dream.
  - comte7092 3y ago
    Death rates have already been trending downward for decades[0] There are a whole host of technologies that exist and have been invented to improve safety, that fall short of full self driving. From vehicle design, to improved road designs, and changes is laws (drunk driving being a major example). Self driving advocates treat it like the messiah that must save us, but there is plenty that can be done to continue to make improvements. > Getting americans to use public transit, requires not only investment in public transit: it requires a cultural shift in how Americans view life. I can see why you would have that opinion, but it is an opinion, not a factual statement. [0] https://en.wikipedia.org/wiki/File:US_traffic_deaths_per_VMT,_VMT,_per_capita,_and_total_annual_deaths.png https://en.wikipedia.org/wiki/File:US_traffic_deaths_per_VMT...
    - Mawr 3y ago
      That's the wrong graph to look at. https://www.nytimes.com/2023/06/27/us/pedestrian-deaths-2022.html https://www.nytimes.com/2023/06/27/us/pedestrian-deaths-2022... https://www.nytimes.com/2022/11/27/upshot/road-deaths-pedestrians-cyclists.html https://www.nytimes.com/2022/11/27/upshot/road-deaths-pedest... Now compare with the Netherlands: https://youtu.be/5aNtsWvNYKE?t=36 https://youtu.be/5aNtsWvNYKE?t=36
lazide 3y ago
Which is in generally cherry-picked areas with very cherry picked weather conditions. And manual take-over when things get ‘weird’. And waymo drives so conservatively to achieve that it drives most drivers around them crazy. No way it would work in Italy, South Korea, India, etc. Still impressive engineering, but not very practical or economic unfortunately,

by adrr 3y ago

Driverless cars are significantly safer than human drivers. https://arxiv.org/ftp/arxiv/papers/2309/2309.01206.pdf https://arxiv.org/ftp/arxiv/papers/2309/2309.01206.pdf

defrost 3y ago

"So far" With less than 4 million miles of pure rider only data by relatively early Waymo rollouts in two cities it can be argued there's room for the claims per million miles to climb as the number of deployments rise.

kilobite88 3y ago
Does it not drive the same roads as human drivers? 4 million miles = 4 million miles.
- comte7092 3y ago
  > Does it not drive the same roads as human drivers? No, human drivers drive on roads beyond selected neighborhoods of Phoenix and San Francisco. It is unknown how generalizable the results of the study would be beyond the conditions experienced on the roads in the areas Waymo had been deployed in. We might see significantly improved safety or we might not. But given the messaging from Waymo, the presumption would seem to be that performance would be worse in other areas/conditions.
  - adrr 3y ago
    Are there more accidents on these roads?
    - comte7092 3y ago
      I am not really sure what your question is asking. More accidents on what roads? If you’re talking about the 99% of roads in the country where Waymo doesn’t currently operate, then I can’t really take your question seriously. We are talking about relative performance of an automated system vs human drivers. The automated system has been specifically tuned to perform well on a certain set of roads. The point is that there is no reason to assume by default that the system would maintain good relative performance under a drastically different set of conditions.
- defrost 3y ago
  4 million miles is proportionally less than the width of a gnat against the history of driver miles in the US across all cities, less than a rounding error. More crucially, these are 4 million entry level service rollout miles .. a time when the company has highly paid engineers paying scrupulous detailed attention to the condition of the cars and the tuning of the automated responses, with jobs dedicated to the avoidance of PR incidents. comte7092 has pointed out other factors - but these are not "typical miles" these are the miles before typical sets in - who can say if the stats will be maintained as sensors age and maintainence becomes an after tought in the manner typical of human driven cars. I hope the figure holds up, I'm commenting here on the apples to oranges issues in the comparison at present.

comte7092 3y ago

I never said that they weren’t/wouldn’t be. But that wasn’t the point I was responding to. The point I was responding to was that we need to address 40,000 traffic deaths per year, to which I said it doesn’t seem to be the most direct route there. Just because self driving cars are safer in some situations doesn’t mean that we will inevitably have that result within a reasonable timeframe and level of investment. Besides the types of things that reduce human caused accidents would lead to benefits for self driving cars: slower speeds, fewer conflicts, etc.

greiskul 3y ago
> a reasonable timeframe and level of investment Compared to the time-frame and investments on which alternatives? It might take 40 years for self driving cars to improve and have enough deployments to significantly reduce the number of accidents, but at least there is a path there and it will happen at some point. Getting americans to use public transit, requires not only investment in public transit: it requires a cultural shift in how Americans view life. A cheaper solution that can never be delivered is not a good solution, it's just a dream.
- comte7092 3y ago
  Death rates have already been trending downward for decades[0] There are a whole host of technologies that exist and have been invented to improve safety, that fall short of full self driving. From vehicle design, to improved road designs, and changes is laws (drunk driving being a major example). Self driving advocates treat it like the messiah that must save us, but there is plenty that can be done to continue to make improvements. > Getting americans to use public transit, requires not only investment in public transit: it requires a cultural shift in how Americans view life. I can see why you would have that opinion, but it is an opinion, not a factual statement. [0] https://en.wikipedia.org/wiki/File:US_traffic_deaths_per_VMT,_VMT,_per_capita,_and_total_annual_deaths.png https://en.wikipedia.org/wiki/File:US_traffic_deaths_per_VMT...
  - Mawr 3y ago
    That's the wrong graph to look at. https://www.nytimes.com/2023/06/27/us/pedestrian-deaths-2022.html https://www.nytimes.com/2023/06/27/us/pedestrian-deaths-2022... https://www.nytimes.com/2022/11/27/upshot/road-deaths-pedestrians-cyclists.html https://www.nytimes.com/2022/11/27/upshot/road-deaths-pedest... Now compare with the Netherlands: https://youtu.be/5aNtsWvNYKE?t=36 https://youtu.be/5aNtsWvNYKE?t=36

lazide 3y ago

Which is in generally cherry-picked areas with very cherry picked weather conditions. And manual take-over when things get ‘weird’. And waymo drives so conservatively to achieve that it drives most drivers around them crazy. No way it would work in Italy, South Korea, India, etc. Still impressive engineering, but not very practical or economic unfortunately,