What?And so do most drivers you encounter next to, in front of, and behind your car every day..Scary though isn't it.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Apple Self-Driving Test Vehicles Doubled Mileage in 2020
- Thread starter MacRumors
- Start date
- Sort by reaction score
There are a lot of people driving around you every day who are on auto pilot.
You're missing the point entirely. The article discusses that Apple doubled the mileage. I'm arguing that Tesla has collected far more mileage of useful data than what Apple has collected.
Sure. I drive on autopilot too and have seen Tesla cars around me on autopilot.
You missed the obvious and serious joke. Drivers all around you not paying attention to what they are doing (hence: auto pilot mode) which can be dangerous.
Your overthinking ruined an obvious joke about bad drivers due to not paying attention. It had nothing to do with actual auto piloted cars. Next time I will save the jokes for someone else who doesn't overthink so much.
If it has nothing to do with autopiloted cars, then the joke "and so do most drivers" in response to billions of miles of autopilot doesn't work because it has nothing to do with billions of autopiloted miles.
It's not overthinking, the joke doesn't connect.
Don't look at disengages as a failure. The system requires human input for training.
Looking at the total miles, Apple is running a VERY modest program.
As secretive as Apple is, why should we assume that the reported numbers are accurate? Why would they tip their hat and broadcast their intentions for a product that may become their biggest product since the iPhone?
the 30-second rule is there as a safety so people don't abuse it since they are public customers not privately hired employees. You seem to think the Apple testing car can ACTUALLY drive itself when its the purpose to learn to drive itself. Tesla's data is unrivalled, its in every single customer car, and in cars all over the world not just one little city with pre-planned routes.
Tesla and Apple have made the same profit selling cars.
The numbers are meaningless without context. Apple could be testing disengagement scenarios, as one example.
Wrong. Even subtracting ev credits, Tesla made a profit.
Can't wait to own a car with a few hundred pounds of cameras on the roof that stick out like a sore thumb. Sexy!
Maybe in a sense, as Waymo and Cruise are way way ahead of Tesla. I'm a Tesla fan and have one on order, but I've watched as many FSD videos as anyone and they disengage *a lot*. Arguably way more than once per 144 miles. That's not to say that it isn't super impressive and won't eventually be great, but it's realistically not anywhere close to the leaders. Their lead is that they have the best cost-effective self driving tech.
I put Tesla in the lead because their tech is low cost, in volume production, and gives them access to huge amounts of training data. Waymo and all the other Lidar-reliant platforms reached a local maxima that will be super hard to leap out of. Cruise only works on select highways that have been hyper-mapped. Tesla and MobilEye are the only platforms using a single unified 360-degree image for sensing the environment.
Tesla still has a ways to go. Path planning is still hardcoded. Once they offload path planning to neural networks, the performance will probably regress a bit as the NNs learn to drive more like humans do. But will quickly improve. And once Tesla's Dojo supercomputer comes online late this year, it's probably game over. I suspect by late 2022 Tesla will be the clear leader in autonomy.
No. It’s autonomous until disengaged. That’s why it had to be reported.
Apple also has millions of miles in virtual simulators where their software is learning.
There's no set criteria for what counts as a disengagement. For example Waymo (or Cruise, I don't remember, probably both) only counts a disengagement if
- The car gives control back to the driver or the driver takes control over the vehicle
- AND the car software calculated that not having given back the control back to the driver would have resulted in an accident where the car would be at fault
Meanwhile companies like Apple that are fully self-funded probably don't want to give away how good their self-driving software is due to it's secrecy. Therefore a company like Apple could be counting literally everything the driver does as a disengagement. And then there's companies like Tesla who are not reporting fully autonomous miles at all because if they would they would never get on par with the low disengagements numbers that Waymo/Cruise are reporting and which would make it seem like they are behind a lot (which they are, current Tesla hardware will never get to L4/L5).
18k miles a year is a joke. That's 50 miles a day of testing. Obviously Apple are not serious about this.
If Apple was serious about true self-driving vehicles and considering the 200 billion dollars it has it its pocket, it can build a small city with roads, traffic signs and everything just for testing purposes.
If Apple was serious about true self-driving vehicles and considering the 200 billion dollars it has it its pocket, it can build a small city with roads, traffic signs and everything just for testing purposes.
And who hasn't? Virtual simulators are garbage. If these were any useful, we would be getting driver licenses with simulators instead of sitting and driving a real car on a real road.
Because if you cheat, you can get banned from the program and if you get banned, you can no longer test your system on public roads. Apart from that, Apple will be putting at stake its reputation and brand image. There is nothing worst than newspaper headline reading "Apple is cheating". In addition, there is high risk of whistleblower revealing what you are doing to the authorities, so you will end up being cought sooner or later and anyone related would be putting his good-paid job in the Silicon Valley at stake for something that isn't worth it really. Bottom line, there is no chance of Apple doing this.