5 ms·

> What use would this type of data be to the average consumer? We can also reverse this question: why would Tesla hide this type of data from their customers s

> What use would this type of data be to the average consumer? We can also reverse this question: why would Tesla hide this type of data from their customers so intensely? One reason that comes to mind on both questions, is that customer's (or someone on their behalf) are able to review if the crash is due to software malfunction. > isn't this otherwise very similar to being in a plane crash and demanding the black box data Most people don't buy (hence own) a plane but buy a ticket with an airplane company instead. The airline company actually does get to access the black box data in case of a crash. So do the authorities.

Navarr 8y ago
> why would Tesla hide this type of data from their customers so intensely? To give them the benefit of the doubt: * Do you consider this diagnostic data to be highly personal, worth protecting heavily? (Does it include GPS?) * Do you trust your engineers to properly verify the identity of a person before granting this potentially highly personal data? * Does your legal department trust your engineers that much? * Do you have the resources/desire to lay out a foundational system of getting that identity verification and granting that data? If you answered "No" to any of those questions, you might very well want to say "We will do it if it is required of us by a court order." While it is unideal, it's not nonsensical.
- danso 8y ago
  But companies like Facebook and Google also have policies that call for only releasing data to government via court order. And both companies allow users to do bulk downloads of their own data.
berdon 8y ago
This isn't reversing the question - it's asking an entirely different question. My question was about the value of the data to a consumer. Presumably, this data is a mixture of location history, sensor/actuator logs (maybe human readable?), and maybe various state data for ML algorithms. Depending on the human "readability" of this data - how could someone without proprietary knowledge find any value in it? That's my question - which is more leading towards "Tesla won't give drivers a means to analyze their crashes" which is a much harder pill to swallow as a Tesla driver, law enforcement officer, etc.
- danso 8y ago
  But Tesla's defense is not that the data is in a proprietary format that can't be decoded. And even if it were, are you suggesting Tesla logs and stores data solely in a non-industry format for which there is not an ETL function that turns it into something less opaque? And what do you base this speculation that Tesla does something so strange? Because Tesla has a machine-learning heavy reputation? Google does far more machine-learning and yet allows us to export data, such as geolocation, in portable JSON format: https://blog.rakam.io/rakam-google-takeout-importer-analyze-your-google-search-and-location-history-data-ba45e4c66b1c https://blog.rakam.io/rakam-google-takeout-importer-analyze-...
  - berdon 8y ago
    "But Tesla's defense is not that the data is in a proprietary format that can't be decoded." I never said it was. I'm not sure why you continue to try to come up with unrelated arguments to push your agenda in a discussion completely unrelated to what you're talking about. "And what do you base this speculation that Tesla does something so strange?" Because educated speculation can lead to positive and constructive discourse? I have a fair amount of experience in the ML and CS field and I could imagine a situation where the data is in fact difficult to analyze. Also, logging geolocation data is a fair amount different than logging neural weights or other types of ML data. Without knowledge of the types and composition of ML algorithms those numbers simply won't help you. Which was my point to begin with.
    - danso 8y ago
      You've proposed a reason for Tesla's reluctance to release data that Tesla itself has not used, so it's you that's bringing up unrelated arguments. Why speculate on the proprietary formats Tesla may or may not use when Tesla doesn't appear to have argued that the data was proprietary?
      - berdon 8y ago
        You're the only one talking about Tesla in this situation. I never mentioned Tesla's reluctance to release data. That's your fabrication. I'm talking about data that's incapable of being analyzed in a meaningful way without contextual / proprietary knowledge of how that data was created, is used, and is stored. In reality, I'd wager that most companies have more of that type of data than not.

Navarr 8y ago

> why would Tesla hide this type of data from their customers so intensely? To give them the benefit of the doubt: * Do you consider this diagnostic data to be highly personal, worth protecting heavily? (Does it include GPS?) * Do you trust your engineers to properly verify the identity of a person before granting this potentially highly personal data? * Does your legal department trust your engineers that much? * Do you have the resources/desire to lay out a foundational system of getting that identity verification and granting that data? If you answered "No" to any of those questions, you might very well want to say "We will do it if it is required of us by a court order." While it is unideal, it's not nonsensical.

danso 8y ago
But companies like Facebook and Google also have policies that call for only releasing data to government via court order. And both companies allow users to do bulk downloads of their own data.

berdon 8y ago

This isn't reversing the question - it's asking an entirely different question. My question was about the value of the data to a consumer. Presumably, this data is a mixture of location history, sensor/actuator logs (maybe human readable?), and maybe various state data for ML algorithms. Depending on the human "readability" of this data - how could someone without proprietary knowledge find any value in it? That's my question - which is more leading towards "Tesla won't give drivers a means to analyze their crashes" which is a much harder pill to swallow as a Tesla driver, law enforcement officer, etc.

danso 8y ago
But Tesla's defense is not that the data is in a proprietary format that can't be decoded. And even if it were, are you suggesting Tesla logs and stores data solely in a non-industry format for which there is not an ETL function that turns it into something less opaque? And what do you base this speculation that Tesla does something so strange? Because Tesla has a machine-learning heavy reputation? Google does far more machine-learning and yet allows us to export data, such as geolocation, in portable JSON format: https://blog.rakam.io/rakam-google-takeout-importer-analyze-your-google-search-and-location-history-data-ba45e4c66b1c https://blog.rakam.io/rakam-google-takeout-importer-analyze-...
- berdon 8y ago
  "But Tesla's defense is not that the data is in a proprietary format that can't be decoded." I never said it was. I'm not sure why you continue to try to come up with unrelated arguments to push your agenda in a discussion completely unrelated to what you're talking about. "And what do you base this speculation that Tesla does something so strange?" Because educated speculation can lead to positive and constructive discourse? I have a fair amount of experience in the ML and CS field and I could imagine a situation where the data is in fact difficult to analyze. Also, logging geolocation data is a fair amount different than logging neural weights or other types of ML data. Without knowledge of the types and composition of ML algorithms those numbers simply won't help you. Which was my point to begin with.
  - danso 8y ago
    You've proposed a reason for Tesla's reluctance to release data that Tesla itself has not used, so it's you that's bringing up unrelated arguments. Why speculate on the proprietary formats Tesla may or may not use when Tesla doesn't appear to have argued that the data was proprietary?
    - berdon 8y ago
      You're the only one talking about Tesla in this situation. I never mentioned Tesla's reluctance to release data. That's your fabrication. I'm talking about data that's incapable of being analyzed in a meaningful way without contextual / proprietary knowledge of how that data was created, is used, and is stored. In reality, I'd wager that most companies have more of that type of data than not.