Waymo seems to be the furthest ahead in my complete outsider opinion. I feel like everyone else is a way distant second.
Assuming the Austin robotaxi service continues to go smoothly, I expect Tesla will leapfrog both of them in the next 12 months. Tesla's cost-effective approach (cameras) combined with the fact that they have already scaled vehicle production positions them really well.
Media (and therefore people who trust it too much) will point out places where each of the services goes wrong, but the reality is that likely all of them are already much safer than human drivers if you define "safety" in terms of severe accidents per million miles.
More than a million people die each year from automobile accidents with humans behind the wheel.
Will sound engineering prevail over brain rot in the C-suite? I am skeptical.
There's no chance this includes Tesla with their disengagements (equivalent to the driver passing out) and even then it's only true under the restricted set of conditions these systems actually operate at compared to human drivers.
I'm guessing that they just stop if they ever encounter a situation where they don't know what to do.
Human drivers don't get the luxury of disengaging and having a more skilled driver take over when they're struggling. If Tesla FSD drives for 100km before overwhelming glare causes the system to disengage, that'll go on record as "100km driven without accident", but when the human driver is blinded by the same glare and ends up in an accident 5km later, that'll go on record as 1 accident per 5km driven for the human driver.
Oh my god that’s terrifying if true. I can think of many situations when driving when slamming on the breaks is the absolute wrong choice. Tesla is pushing this out way before it’s safe enough to operate in public.
1. Keep driving in some way
2. Stop
Which one do you think Waymo (or any other system) does when it doesn't understand a situation?
Besides whether the video is biased or not, are we conflating comfort & safety? When a robotaxi jiggles the wheel and makes the car shake, that's clearly terrible for comfort and perhaps brings to mind an inexperienced driver, but that is a personification. Programmers who have ever dealt with debouncing might recognize that it's likely something that could use smoothing to aid with comfort, but not necessarily unsafe.
In any case, Apollo Go is ahead in number of vehicles, cities, rides, and probably also miles. Not that it really matters. Demand is going to lead supply in autonomy for many years to come. These services are not actually going to be in real competition with each other for quite some time.
I was in China (Beijing) a couple of months ago yet didn't have a chance to ride a driverless car there because they were only running them in the deep suburbs (Changping near Lenovo I think?). This is quite different from my experience taking Waymo, where it served the central city just fine (although I'm not sure how Waymo would handle a narrow Hutong alley way like one of the taxi's I took did).
Once I saw the part where the car was being blamed for the pickup location chosen by the user, I really struggle to believe anything that's being said. Felt like they wanted to say it's not good right from the start.
This started with me pointing out that the scale of Apollo Go is greater than Waymo, but then you made it about safety, which is quite a different (but still interesting) topic. Then there's this video that has been brought up in the context of safety, but seems more related to ride comfort. Ride comfort is also interesting, but different than safety.
I'm not sure we can actually have a reasonable conversation without objectively separating such topics.
I haven't looked at what the service area is, but I wouldn't assume that a downtown service area is somehow better or worse than a suburban service area, especially not without defining "better". A suburban-focused service seems like it might actually be a lot more useful than a downtown-focused service.
Back to the original point, you easily see the data yourself. The scale of Apollo Go is slightly larger than Waymo on most metrics that would seem to matter; vehicles, rides, miles, etc.
- Elon, probably
There's a high floor to Waymo's costs though, and therefore a lower ceiling to their profits. Tesla's aim was always to make it (way) more scalable. If they can match Waymo's performance at Tesla's costs, Waymo goes the way of the dodo pretty quick.