Let's face it, "10x" has no objective or empirical justification, it is just an arbitrary standard based on a large round number. "Parity" can be justified on the basis of net lives saved. Any local traffic or regulatory authority that implemented some new policy or built new traffic management infrastructure that could report a rapid shift to even "25 percent fewer serious accidents per vehicle mile" would be crowing about it with the leaders in charge putting that accomplishment on the top line of their CV and preparing to make their acceptance speeches at the international traffic regulators awards banquet. Imagine telling a local traffic authority that they were not authorized to make any unambiguously safety-improving and cost-benefit test-passing changes at all - such as adjusting stoplight timing or switching from a right angle intersection to a roundabout - unless that change made things at least 10x safer than the status quo. Literally nothing would ever happen, it's a recipe for paralysis and a permanent freezing of everything in the same place forever. It is almost by luck that this standard can even seem remotely reasonable as some AV technology has already plausibly met the 10x threshold, but that doesn't justify the threshold so much as start what most perspicacious people can already see coming which is to gradually *unjustify* tolerance of human driving.
...and I wonder if AV firms would go so far as to challenge investment in safety-related road improvements (e.g. converting an intersection to a roundabout) on the premise that it would change their baseline and make the "10x safer" bar more difficult to achieve. To go a bit further, if there is a municipality (perhaps in Sweden) that has gone full "Vision Zero" and made their infrastructure (and support systems, like driver training and testing) as safe as it could reasonably be, it would be impossible for an AV operator to then go 10 x further from that baseline; consequently they would effectively be precluded from working there. I therefore suggest that the "10x safer" bar (if that particular value has some public / political traction - I agree that "why not 8x or 6x?" is a legitimate question) should be a broader goal and support should be given to ALL measures that achieve it. Don't just sit back and let technology take care of it. Demand management, vehicle restrictions, driver training, cyclist licensing, infrastructure changes, better sidewalks, technology improvements, and more all have a role to play in making the world safer. So AV firms can contribute to non-AV activities that help get to the overall goal, and maybe AV itself is only half or less of the solution.
I also wonder about the baseline - is it a 2026 snapshot, or a continuously moving target? I note that safety risk and related violations are notoriously poorly documented; for example I, like most people, routinely exceed the posted speed limit while driving (the 85th percentile speed on a road is often well above the posted limit) which theoretically constitutes an increased safety risk in terms of reduced reaction time and increased impact should I hit something. Yet "collisions" - the outcome - rather than "speed" - the risk - are used as the shorthand for the AV "safety improvement" baseline discussed here. Of course, in a full AV world, speed limits become irrelevant, so a 10x improvement in aligning operating speed with posted speed limits is not really a "safety improvement".
It's an interesting question as to whether there even exists any other example of a policy that will not allow a demonstrably safer alternative "unless it is at least X times safer than the status quo." Every attempt to compose that sentence in any other field of safety strikes one as immediately absurd and ridiculous. FFSS talks about copying regimes from other sectors but they have failed to do so with their one big suggestion with an X factor they have pulled out of thin air that applies to no other safety circumstance in any other industry, and for good reason!
Also, I've got to say - while this annoying practice is unfortunately increasingly common - it's still pretty lame for any organization to come out with a white paper one can't simply access and download from the main website without giving them your email to subscribe to their newsletter (and to give to who knows which other "partners" down the line.) As the kids might say, it's immediately sus and cringe when organizations do that.
As I wrote above, the larger context here is the presence a large and motivated group of people who would be happy to shut down this technology irrespective of its safety impacts, just because it 'kills jobs' or 'empowers tech companies' or whatnot. A group that includes the next PM of the UK, so they are no mere lunatic fringe.
In this environment, holding a high bar on safety is the best way to win the uncommitted to your side. And it's not that high a bar: Waymo has already achieved it!
Well said and I agree. I suspect, however, that meeting even unreasonably high bars (15x? 20x?) on safety wouldn't be enough as the goal posts will keep moving for people with those ulterior motives who are determined to keep this tech off the road no matter what.
I must say I find it a little mysterious that with all that Alphabet money and government-influencing prowess behind it, the Waymo lobbying efforts to date have proven so apparently weak, amateurish, and inadequate. Maybe one day we'll learn the true inside story of how they fumbled that aspect of the rollout strategy. I suspect they simply seriously underestimated the political resistance that would emerge and by the time they got their act together it had already acquired too much cultural momentum. But I really have no idea.
The 10x framework makes more sense to me if it's just "this is what we think is politically feasible" because, from a purely moral point of view, I don't understand why FSS or anyone would adamantly oppose a technology that results in 9x, or even 2x, fewer fatalities than current human driving.
I would go further, and argue that deployment should be legal at a bit below parity (so long as the companies are strictly liable for damage). There's a long gap between being able to deploy at all and universal adoption, having real world data helps a lot with increasing safety, and the liability ensures the companies are heavily incentivised to keep working on improving their safety. We let below-average human drivers keep driving so long as they're not consistently dangerous, using the same bar for AV seems reasonable. (You mention allowing trials at only 2x or 5x safety, but supervised trials just seem like allowing it with extra regulatory burden).
I think this broadly just underpriced the negatives of regulation. Even a perfectly complient AV maker will have to spend years just doing paperwork, causing many deaths and delayed deployment (not to mention harming market competition). The comparison to the nuclear safety standard - which effectively killed north American nuclear power - is instructive here.
If everyone was enthusiastic about robotaxis, I'd be more sanguine about a light-touch regulatory policy.
But there are lots of people opposed to this technology, like the next PM of the UK for instance, and being cavalier about safety will just give those opponents ammunition.
I'm all for saving lives on net. On the longest time horizon, that means proceeding at a measured pace today, not risking a catastrophic setback, like the one that effectively killed Cruise:
I see the argument, but I think the pro-safety people will always ask for more safety no matter how much you offer, so it's better to start at what you actually want than to partly concede in advance in the hopes of appeasing them. I think the cruise example also shows why it's incumbent on outsiders to be less conciliatory - it's much easier for me to say "no this is too much safetyist regulation" than for an AV company, which would be putting a target on its back if it said this no matter if it was right or wrong.
Let's face it, "10x" has no objective or empirical justification, it is just an arbitrary standard based on a large round number. "Parity" can be justified on the basis of net lives saved. Any local traffic or regulatory authority that implemented some new policy or built new traffic management infrastructure that could report a rapid shift to even "25 percent fewer serious accidents per vehicle mile" would be crowing about it with the leaders in charge putting that accomplishment on the top line of their CV and preparing to make their acceptance speeches at the international traffic regulators awards banquet. Imagine telling a local traffic authority that they were not authorized to make any unambiguously safety-improving and cost-benefit test-passing changes at all - such as adjusting stoplight timing or switching from a right angle intersection to a roundabout - unless that change made things at least 10x safer than the status quo. Literally nothing would ever happen, it's a recipe for paralysis and a permanent freezing of everything in the same place forever. It is almost by luck that this standard can even seem remotely reasonable as some AV technology has already plausibly met the 10x threshold, but that doesn't justify the threshold so much as start what most perspicacious people can already see coming which is to gradually *unjustify* tolerance of human driving.
...and I wonder if AV firms would go so far as to challenge investment in safety-related road improvements (e.g. converting an intersection to a roundabout) on the premise that it would change their baseline and make the "10x safer" bar more difficult to achieve. To go a bit further, if there is a municipality (perhaps in Sweden) that has gone full "Vision Zero" and made their infrastructure (and support systems, like driver training and testing) as safe as it could reasonably be, it would be impossible for an AV operator to then go 10 x further from that baseline; consequently they would effectively be precluded from working there. I therefore suggest that the "10x safer" bar (if that particular value has some public / political traction - I agree that "why not 8x or 6x?" is a legitimate question) should be a broader goal and support should be given to ALL measures that achieve it. Don't just sit back and let technology take care of it. Demand management, vehicle restrictions, driver training, cyclist licensing, infrastructure changes, better sidewalks, technology improvements, and more all have a role to play in making the world safer. So AV firms can contribute to non-AV activities that help get to the overall goal, and maybe AV itself is only half or less of the solution.
I also wonder about the baseline - is it a 2026 snapshot, or a continuously moving target? I note that safety risk and related violations are notoriously poorly documented; for example I, like most people, routinely exceed the posted speed limit while driving (the 85th percentile speed on a road is often well above the posted limit) which theoretically constitutes an increased safety risk in terms of reduced reaction time and increased impact should I hit something. Yet "collisions" - the outcome - rather than "speed" - the risk - are used as the shorthand for the AV "safety improvement" baseline discussed here. Of course, in a full AV world, speed limits become irrelevant, so a 10x improvement in aligning operating speed with posted speed limits is not really a "safety improvement".
I digress....
It's an interesting question as to whether there even exists any other example of a policy that will not allow a demonstrably safer alternative "unless it is at least X times safer than the status quo." Every attempt to compose that sentence in any other field of safety strikes one as immediately absurd and ridiculous. FFSS talks about copying regimes from other sectors but they have failed to do so with their one big suggestion with an X factor they have pulled out of thin air that applies to no other safety circumstance in any other industry, and for good reason!
Also, I've got to say - while this annoying practice is unfortunately increasingly common - it's still pretty lame for any organization to come out with a white paper one can't simply access and download from the main website without giving them your email to subscribe to their newsletter (and to give to who knows which other "partners" down the line.) As the kids might say, it's immediately sus and cringe when organizations do that.
As I wrote above, the larger context here is the presence a large and motivated group of people who would be happy to shut down this technology irrespective of its safety impacts, just because it 'kills jobs' or 'empowers tech companies' or whatnot. A group that includes the next PM of the UK, so they are no mere lunatic fringe.
In this environment, holding a high bar on safety is the best way to win the uncommitted to your side. And it's not that high a bar: Waymo has already achieved it!
Well said and I agree. I suspect, however, that meeting even unreasonably high bars (15x? 20x?) on safety wouldn't be enough as the goal posts will keep moving for people with those ulterior motives who are determined to keep this tech off the road no matter what.
I must say I find it a little mysterious that with all that Alphabet money and government-influencing prowess behind it, the Waymo lobbying efforts to date have proven so apparently weak, amateurish, and inadequate. Maybe one day we'll learn the true inside story of how they fumbled that aspect of the rollout strategy. I suspect they simply seriously underestimated the political resistance that would emerge and by the time they got their act together it had already acquired too much cultural momentum. But I really have no idea.
The 10x framework makes more sense to me if it's just "this is what we think is politically feasible" because, from a purely moral point of view, I don't understand why FSS or anyone would adamantly oppose a technology that results in 9x, or even 2x, fewer fatalities than current human driving.
I would go further, and argue that deployment should be legal at a bit below parity (so long as the companies are strictly liable for damage). There's a long gap between being able to deploy at all and universal adoption, having real world data helps a lot with increasing safety, and the liability ensures the companies are heavily incentivised to keep working on improving their safety. We let below-average human drivers keep driving so long as they're not consistently dangerous, using the same bar for AV seems reasonable. (You mention allowing trials at only 2x or 5x safety, but supervised trials just seem like allowing it with extra regulatory burden).
I think this broadly just underpriced the negatives of regulation. Even a perfectly complient AV maker will have to spend years just doing paperwork, causing many deaths and delayed deployment (not to mention harming market competition). The comparison to the nuclear safety standard - which effectively killed north American nuclear power - is instructive here.
If everyone was enthusiastic about robotaxis, I'd be more sanguine about a light-touch regulatory policy.
But there are lots of people opposed to this technology, like the next PM of the UK for instance, and being cavalier about safety will just give those opponents ammunition.
I'm all for saving lives on net. On the longest time horizon, that means proceeding at a measured pace today, not risking a catastrophic setback, like the one that effectively killed Cruise:
https://www.changinglanesnewsletter.com/i/152986950/the-rise-and-fall-of-cruise
I see the argument, but I think the pro-safety people will always ask for more safety no matter how much you offer, so it's better to start at what you actually want than to partly concede in advance in the hopes of appeasing them. I think the cruise example also shows why it's incumbent on outsiders to be less conciliatory - it's much easier for me to say "no this is too much safetyist regulation" than for an AV company, which would be putting a target on its back if it said this no matter if it was right or wrong.