Condition for resolving this market:
At least 10000 commercial rides have happened
Tesla's robotaxi is available to the general public in the US. So not behind a waitlist.
The robotaxis are fully driverless and FSD.
The type of car doesn’t matter nor if it has a steering wheel. Just needs to be self driving without any humans in the chain.
Rides should happen from a ride hailing service.
1,000@DavidFWatson I'd say the critical feature is whether it's self-driving. Who really cares whether it has a steering wheel at that point, if it's just some backup thing that barely ever gets used?
@makoyass I'm concerned that it could become a question of words rather than reality. The current Teslas literally have a feature called "full self driving", and yet we probably agree that they do not qualify as 'self-driving'.
Perhaps it could be based on whether the 10k commercial rides involved the car driving empty to pick up the passenger?
@DavidFWatson We can certainly easily agree about that :/ I don't think it's really difficult to call whether it's an autonomous service. I think the 10k is a decent criterion.
@makoyass I had a small bet with a friend regarding Waymo rolling out self driving taxis to all of SF. Now that it's clearly 'YES' he argues that it's not because the cars 'aren't actually self-driving' given that they can request assistance from humans at waymo if they get stuck.
"Are thousands of cars driving around with no one in them" is a nice easy benchmark that no one will argue about later.
@DavidFWatson so, you're saying remote-controlled would qualify as self-driving? I'd strongly disagree.
@deagol not completely remote-controlled, but if it's very rarely remote-controlled when there's a problem that seems fine? it's how i would intuitively take it to mean. tesla having the ability to override the ai doesn't mean the ai isn't driving, or wtv
@deagol Latency and connectivity are such that you can't reliably remote control a car over long distances. But there's also no reason to build such a system, because it'd be uneconomical to operate.
Worrying about distinguishing that end of the spectrum doesn't seem valuable to me.
On the other hand, I'm interested in a criteria that handles the other side of the spectrum cleanly: cars that are never helped with remote assistance vs cars that are helped remotely once per a hundred trips.
@deagol you can have a viable self-driving car business with the occasional intervention. But if you're more interested in the technology then the economic implications you won't be impressed by self-driving that requires interventions. It's also going to be difficult to call exactly how much intervention would be allowed.
@deagol why care about it for anything other than the economic implications!
If you built a truly self driving car but it was non viable because the computation absolutely required 100kw to operate, no one would care!
@DavidFWatson The steering wheel doesn’t matter. Just self driving autonomously without any humans. Rides should be in function of a ride hail service.
@RutgerDeMaeyer Is the 10000 rides limit the number that occur without any intervention? So for example if there are 11000 rides in 2025 and it is estimated that 2% have human intervention then do you calculate 11000*0.98 >10000 so it resolves yes?
A second possibility is that "Just needs to be self driving without any humans in the chain." means that the 2% intervention rate means that none of them count? The trouble with this interpretation is that if 2% is too high, then what is ok? If intervention rate is 1 in 5000 or 1 in 15000 the service is essentially working autonomously with rare intervention. So it seems it should resolve yes with some low intervention rate.
It is also possible you wouldn't want to count 22000 rides with a 50% intervention rate. So to exclude this you might want to set a maximum intervention rate. e.g. 10000 rides with a maximum intervention rate of 2%
@ChristopherRandles The relevant standard is SAE level 4 autonomy, like Waymo has. No human in the driver's seat but the car can stop and get guidance from a remote human if needed (like when there are firetrucks in the road, cops directing traffic, something where some human judgment is wanted).
If there's a human backup driver in the driver's seat ready to intervene in real time with the car in motion, that shouldn't count, regardless of how rare the interventions are.
@professionalgambler Yes the car doesn’t matter it just needs to be self driving and the passenger doesn’t touch the steering wheel. The rides I mentioned should be from a uber like service that is available. (Ride hailing).
@DavidFWatson No, robotaxi refers to the service not the car. The car doesn’t matter. The question is more about general availability.
