As an outsider, the reason why Waymo is way ahead is clear. Extremely talented team, access to Google infrastructure and expertise, deep pockets and a culture of prioritizing safety over everything.
This ignores the cost of sending data wirelessly over cellular networks so remote operators can see the vehicle, the cost of maintaining the software and hardware on the vehicle itself, the cost of maintaining the vehicle fleet, the cost of creating and operating a consumer-facing support system (!!! for Google), the cost of dealing with the enormous amounts of data those vehicles create (even for Google this is not trivial), the cost of maintaining the special HD maps necessary for the vehicles, not to mention the sunk cost of billions spent developing it.
That's not to say you shouldn't dream big (it's a moonshot after all). But there are plenty of reasons to think it won't be viable even if they can solve the technical challenges, and that much still isn't even clear yet.
The phrase "value of full autonomy" is excluding fleet costs on purpose. It's about drivers.
Customer support is also going to be far far fewer than one person per vehicle.
The enormous amounts of data? If it's not valuable they can just discard it! Having sensors attached to something doesn't obligate you to store it forever.
Everyone knows the software/configuration costs are immense here. But that's the lion's share of the difficulty, and there's no reason to act like minor hurdles are bigger than they are.
> The enormous amounts of data? If it's not valuable they can just discard it!
I'm curious what the legal requirements for this will be. But I imagine they'd want to hold on to data of driving scenarios for at least a month, in case they get accused of wrongdoing by other drivers. If they had no data to back up their case with all those sensors, it would look awfully suspicious and essentially one witness against nobody - so they'd have to hold onto the data for however long the legal teams deem is okay.
This may sound crazy, but it's already happening at the scale of testing with just a few dozen cars.
They can be accused of a crash even if there isn't a crash though. And if they have no data to back up their side of the story, why would any court believe them? In other words - people can just randomly accuse them of hit-and-runs and they'd have nothing to say otherwise.
I don't see how it's any different from accusing random people of a hit and run.
Though enough data to disprove a hit and run wouldn't actually take up very much space. Medium-resolution camera views and some acceleration data? Sure, pop a single SSD in there and it'll hold more than a month's logs.
Most importantly, their fleet could be made obsolete by personal ownership of assisted driving vehicles. So by any of car manufacturers... (Better version of Tesla autopilot for instance.)
