4 ms·

Self-Driving Cars: Waymo Beats Tesla by Picking the Right Target

allears 6y ago
Not really an article, just a bunch of opinions strung together with little in the way of explanation or background. Waymo and Tesla have different approaches, complete with different strengths and weaknesses. There is nobody, including this author, who knows in advance how self-driving will evolve.
- xkjkls 6y ago
  High uncertainty doesn't mean predictions are useless.
maxharris 6y ago
I don't understand what they mean by "beats Tesla". I can buy a Tesla today with the FSD option, and get the FSD beta as a download, as evidenced by all the videos of ordinary people using it on Twitter. I can't buy a Waymo. I can't even buy a ride on one where I live in Southern California.
- xkjkls 6y ago
  What they advertise as "Full Self Driving" is very far from it.
- johnnyb_61820 6y ago
  The FSD "beta" is ridiculous. They act like FSD is done except for the edge cases. The whole problem of FSD is the edge cases. And, on public roads, edge cases are encountered frequently. This is a problem of targeting Level 5. There's no actual way to do it. It is counter to any other way of automating anything. Automation occurs by regularizing the environment. Every time. Level 4 focuses on finding or creating highly predictable environments. This is what allows for automation. By regularizing the environment you can structurally eliminate the edge cases. It works like a virtual rail system. That works. That's the future. That's the way that automation itself actually happens in the real world.
  - maxharris 6y ago
    Why is resolving those edge cases anything but a matter of collecting enough live data from the actual cars, as they are doing right now? I think it will be pretty easy for Level 5 to be an improvement over human drivers that get tired. And let's not forget that some number of people decide to drive drunk and high, along with a whole host of other factors. In total, somewhere around 35k people die per year in auto accidents. So I think FSD has a pretty low bar to beat.

Not really an article, just a bunch of opinions strung together with little in the way of explanation or background. Waymo and Tesla have different approaches, complete with different strengths and weaknesses. There is nobody, including this author, who knows in advance how self-driving will evolve.

xkjkls 6y ago
High uncertainty doesn't mean predictions are useless.

maxharris 6y ago

I don't understand what they mean by "beats Tesla". I can buy a Tesla today with the FSD option, and get the FSD beta as a download, as evidenced by all the videos of ordinary people using it on Twitter. I can't buy a Waymo. I can't even buy a ride on one where I live in Southern California.

xkjkls 6y ago
What they advertise as "Full Self Driving" is very far from it.
johnnyb_61820 6y ago
The FSD "beta" is ridiculous. They act like FSD is done except for the edge cases. The whole problem of FSD is the edge cases. And, on public roads, edge cases are encountered frequently. This is a problem of targeting Level 5. There's no actual way to do it. It is counter to any other way of automating anything. Automation occurs by regularizing the environment. Every time. Level 4 focuses on finding or creating highly predictable environments. This is what allows for automation. By regularizing the environment you can structurally eliminate the edge cases. It works like a virtual rail system. That works. That's the future. That's the way that automation itself actually happens in the real world.
- maxharris 6y ago
  Why is resolving those edge cases anything but a matter of collecting enough live data from the actual cars, as they are doing right now? I think it will be pretty easy for Level 5 to be an improvement over human drivers that get tired. And let's not forget that some number of people decide to drive drunk and high, along with a whole host of other factors. In total, somewhere around 35k people die per year in auto accidents. So I think FSD has a pretty low bar to beat.