4 ms·

You'd want the car to avoid mistaking reflective surfaces for tunnels. Or not seeing a truck turned sideways due to lack of AI training data. Plus your car shou

You'd want the car to avoid mistaking reflective surfaces for tunnels. Or not seeing a truck turned sideways due to lack of AI training data. Plus your car should make it easy for you to do the right & safe thing. "The nation’s top safety investigator slammed Tesla on Tuesday for failing to take adequate measures to prevent “foreseeable abuse” of its Autopilot driver-assistance technology" https://www.latimes.com/business/story/2020-02-25/tesla-autopilot-crash-hearing https://www.latimes.com/business/story/2020-02-25/tesla-auto...

roenxi 7y ago
For one that is changing the subject. For two; proving that a Tesla autopilot caused an accident and the regulators are grumpy isn't evidence of anything. Humans frequently have accidents and regulators are typically grumpy. And the evidence in that article is that the car is basically working on autopilot. So the brand name is quite reasonable. If you expect any transport system not to have accidents you are crazy - all modes of transport have accidents from time to time. Also; a tired point; but Tesla's autopilot is going to learn from each fatal accident. That is a massive difference from humans; who stop learning after fatal accidents. Fatal accidents are not going to be a major long-term problem here which stands in sharp contrast to human drivers. It isn't a particularly bad sign as it means Tesla's autopilot will have learned a valuable lesson due to someone-who-is-not-me's misfortune. I'd happily pay money for that. Deaths from bad human driving are meaningless tragedies. Deaths in a new but engineer-able system are extremely meaningful tragedies and are almost good news for everyone who isn't immediately involved.
- jacobush 7y ago
  Yes, but how much can a frog really learn. They need to "evolve" new sensors and better onboard processing packages, but these things are expensive.
- fxtentacle 7y ago
  My prediction is that we'll need GPUs with 60+ GB of memory before self-driving becomes more plausible. So that'll be a completely different hardware than what Tesla currently sells as their "self-driving hardware" package.
  - roenxi 7y ago
    > My prediction is that we'll need GPUs with 60+ GB of memory before self-driving becomes more plausible. Somebody has already linked an article showing that people in these cars are taking their hands off the wheel and playing videogames while 'driving'. These cars are already self-driving by a reasonable standard, and is certainly fairly described as an autopilot system. Your standards aren't normal.
    - close04 7y ago
      > These cars are already self-driving by a reasonable standard It's not that others have too high standards, it's that you have low standards and they make basically every mundane thing, that you use every, day a "self-whatever" thing.
    - jsight 7y ago
      > Somebody has already linked an article showing that people in these cars are taking their hands off the wheel and playing videogames while 'driving'. That guy died in a car crash while doing that.
      - roenxi 7y ago
        Yeah? It isn't reasonable to expect no accidents; so that doesn't really tell us very much. It might be part of a broader pattern or it might not be. But it is evidence that there will be a lot of Tesla drivers out there leaving the car on autopilot which is pretty good support for naming the system 'autopilot'.
- jsight 7y ago
  > proving that a Tesla autopilot caused an accident and the regulators are grumpy isn't evidence of anything. Humans frequently have accidents and regulators are typically grumpy. Don't be shocked if they[1] do write guidelines for driver attention monitoring that Tesla has to adhere to. Also, don't be shocked if they apply across the board, even to cars that do not have autopilot like features. [1] By "they" in this case, I do not mean the writers of that report, as that was the NTSB. The NHTSA will likely investigate the concepts more holistically.

roenxi 7y ago

For one that is changing the subject. For two; proving that a Tesla autopilot caused an accident and the regulators are grumpy isn't evidence of anything. Humans frequently have accidents and regulators are typically grumpy. And the evidence in that article is that the car is basically working on autopilot. So the brand name is quite reasonable. If you expect any transport system not to have accidents you are crazy - all modes of transport have accidents from time to time. Also; a tired point; but Tesla's autopilot is going to learn from each fatal accident. That is a massive difference from humans; who stop learning after fatal accidents. Fatal accidents are not going to be a major long-term problem here which stands in sharp contrast to human drivers. It isn't a particularly bad sign as it means Tesla's autopilot will have learned a valuable lesson due to someone-who-is-not-me's misfortune. I'd happily pay money for that. Deaths from bad human driving are meaningless tragedies. Deaths in a new but engineer-able system are extremely meaningful tragedies and are almost good news for everyone who isn't immediately involved.

jacobush 7y ago
Yes, but how much can a frog really learn. They need to "evolve" new sensors and better onboard processing packages, but these things are expensive.
fxtentacle 7y ago
My prediction is that we'll need GPUs with 60+ GB of memory before self-driving becomes more plausible. So that'll be a completely different hardware than what Tesla currently sells as their "self-driving hardware" package.
- roenxi 7y ago
  > My prediction is that we'll need GPUs with 60+ GB of memory before self-driving becomes more plausible. Somebody has already linked an article showing that people in these cars are taking their hands off the wheel and playing videogames while 'driving'. These cars are already self-driving by a reasonable standard, and is certainly fairly described as an autopilot system. Your standards aren't normal.
  - close04 7y ago
    > These cars are already self-driving by a reasonable standard It's not that others have too high standards, it's that you have low standards and they make basically every mundane thing, that you use every, day a "self-whatever" thing.
  - jsight 7y ago
    > Somebody has already linked an article showing that people in these cars are taking their hands off the wheel and playing videogames while 'driving'. That guy died in a car crash while doing that.
    - roenxi 7y ago
      Yeah? It isn't reasonable to expect no accidents; so that doesn't really tell us very much. It might be part of a broader pattern or it might not be. But it is evidence that there will be a lot of Tesla drivers out there leaving the car on autopilot which is pretty good support for naming the system 'autopilot'.
jsight 7y ago
> proving that a Tesla autopilot caused an accident and the regulators are grumpy isn't evidence of anything. Humans frequently have accidents and regulators are typically grumpy. Don't be shocked if they[1] do write guidelines for driver attention monitoring that Tesla has to adhere to. Also, don't be shocked if they apply across the board, even to cars that do not have autopilot like features. [1] By "they" in this case, I do not mean the writers of that report, as that was the NTSB. The NHTSA will likely investigate the concepts more holistically.