Clarifying questions before I bet:
The John Carmack & Jeff Atwood wager is about level 5 autonomy and this is about level 4, right? Waymo is level 4, meaning no human in the driver's seat nor remotely in control when the car is driving but the car may have restrictions on where it can go and in what conditions and may require human input to make confusing decisions. (I think the human only gives that input when the car is stopped. So the human is not in the loop in real time when the car is in motion.) In any case, I predict that level 4 will be kind of ubiquitous before any city has level 5. As of late 2024, level 5 doesn't exist at all.
Does it count as "generally available" if there's a wait list?
What Chinese cities currently count?
Same story not paywalled, on Waymo's blog: https://waymo.com/blog/2024/09/waymo-and-uber-expand-partnership/
So just in the US we're about to have
Phoenix
San Francisco
Los Angeles
Austin
Atlanta
But Austin and Atlanta won't be available to the public until "early 2025" and this market is about 2025 January 1.
@ScottBlanchard if you read this piece, you'll notice it links to another Waymo release. While Zoox and others may be extensively using remote control, Waymo (like Apollo Go) is not:
https://waymo.com/blog/2024/05/fleet-response/
Having a remote helpline that inquires to the city which lanes are meant to be blocked off by cones is not a "remote human driver". If human Uber drivers were using a service like this, it would be hilarious to describe it as such.
@benshindel Mate - As I said, I no longer have a dog in this fight. I did read the blog post and I don't think you are grasping why they are contorting the language the way they are. Again, I exited the market already. I absolutely respect your right to your deluded understanding of what these companies are passing off as "fully autonomous". I see a ton of motivated reasoning in my line of work and have learned to ignore it. Please consider me as having no interest in this market and how it resolves. Thank you and vaya con dios!
@ScottBlanchard I also currently have no shares, so I’m not sure what my motivation would be. Your motivation to hold onto the last vestiges of human autonomy seems pretty understandable though.
@ScottBlanchard more so than Waymo? They also use remote drivers to get out of tricky situations, but I think they clearly count for the resolution of this question.
@HenriThunberg Well - your resolution says “no test, backup or remote human driver”, so if services using remote drivers count, perhaps you shouldn’t have written that they don’t when writing the resolution rules.
@HenriThunberg In any event, I exited the question so I no longer have a dog in the fight.
@ScottBlanchard not my question, but I'm exiting my YES based on agreeing with you that resolution criteria are tricky and ambiguous.
@HenriThunberg My apologies - I assumed that as you replied to me it was your question. My mistake entirely and once again I am relearning the lesson of the dangers of making assumptions.
@benshindel Your understanding is wrong. As with Waymo and I suspect every ‘autonomous’ car out there, there is extensive human intervention. The companies don’t highlight (or even acknowledge) this normally, but Waymo had to release a ton of data in response to the CA investigation and the amount of human intervention is shocking. The Chinese companies have also admitted that there is some remote intervention, but I see no hard data on how much. Given Waymo’s data, though, and the relatively lower cost of human monitors in China along with the higher stakes of failure, my baseline assumption is that the human component is even larger in China.
@intellectronica you probably want to weigh in here, for the sake of the market?
I think your spirit is that adding new Waymo cities count, but if Scott's info above stands uncontested then it's bery hard to know what to bet on for people:)
Perhaps closing market and relaxing the criteria for autonomous a bit?
@ScottBlanchard Your assumption is mistaken. Apollo Go is fully autonomous. The presence of human monitors (as with Waymo) is not relevant to this question. Monitors are not “remote drivers”!
More evidence for “commercial licenses” in Wuhan, Shenzhen, Beijing for Apollo Go.
This article says commercial licenses are starting this month in Shanghai (although I’m pretty sure it’s just bad journalism since Apollo Go already had this for over a year, but I think it’s referring to a new batch of licenses for Apollo Go AND three other companies)
https://finance.yahoo.com/news/shanghai-put-driverless-robotaxis-roads-093000308.html
I think it’s pretty clear at this point that Apollo Go does have commercial licenses in the cities that its website SAYS it’s operating commercially in. I don’t know how feasible it is to find English language coverage for every city though?
That being said,
Phoenix, Mesa, SF, LA, Wuhan, Beijing, Shanghai, Shenzhen, Chongqing, Guangzhou makes 10. Plus another 4+ if you believe Apollo Go’s website! And we still have 5 months to go.
Last week Ben Shindel claimed that this had already happened.
I found a post from Reddit:
“I had a co-worker visit Wuhan recently and he was able to try it without knowing someone.”
“it’s a blended fleet, meaning that depending on your start/end point you’ll either get a driverless car or a driver red car. He couldn’t exactly tell, but the driverless territory appears to be about a quarter of the “full” service area, which is itself ~75% of Wuhan.
He did a driverless ride that was in a quiet part of town and generally went well, the pickup was wonky as pickup spots are sparse and cars really a ick to them, so it double parked unnecessarily. He did a delivered ride in a denser area, and there were frequent disengages as progress was slow otherwise.”
Is this is all true, would you consider this as a city that satisfies the criteria?
@intellectronica are you willing to answer the question above? It will help me calibrate my wagering here.
@JimAusman hey! I think I've engaged quite fairly on this market, presenting long threads with evidence and not making unverifiable claims. Waitlists (supply constraints) are not precluded by this market explicitly, and many people I know have ridden driverless cars in SF and Phx. I assumed that since there was no waitlist in Phx, it had been removed prior in SF, but it turned out SF was slower. But 300k ppl had ALREADY ridden Waymo driverless cars in SF (larger than the entire city pop required by this market).
That’s a tough call. I think if there’s a good chance that you’ll get a human driver that would have to be a no. If it’s a rare exception I’d argue we should include this city. What would you suggest are sensible rules in this case? I’m open to advice on what’s a helpful interpretation in the spirit of this bet. What I really care about is is "robotaxi tech is widely deployed in non-experimental setting". The specific rules are just to make things measurable and rational.