4 ms·

It seems like there's a large volume of information in the logs. I mean, if Tesla can tell whether the driver has his hands on the wheels there's probably a ton

It seems like there's a large volume of information in the logs. I mean, if Tesla can tell whether the driver has his hands on the wheels there's probably a ton of other details. It makes me wonder how practical it would be to use these logs. I hear about so-called "black boxes" in cars (not just Tesla), but where are the tools to decipher the data? How much can you trust that the logs reflect reality?

hashkb 10y ago
It might be difficult, but that shouldn't impact the right of drivers (if any) to try.
- crispyambulance 10y ago
  I agree, and it's not just drivers that would be able to make use of such "black box" log data but also insurance companies and law enforcement. The thing is it would require A LOT of cooperation from Tesla. Once Tesla starts sharing logs, they have to support it and can no longer change it at a "whim" for internal purposes. Perhaps most importantly, Tesla would also have to defend the accuracy of the logs when they go against other observations. Just because something is in a log, doesn't mean it correct, accurate or true. I can understand why they would exert a mighty effort to avoid giving out logs.
mikeash 10y ago
I'm skeptical of the "hands on the wheel" thing. They can't actually sense hands on the wheel, what they sense is torque on the steering wheel. I often have my car nag me to put my hands on the wheel when I already have them there, and I'm just not applying enough torque with them. It's possible that the threshold for the nag is just higher, and that the torque sensor is so sensitive there's no way it could possibly have a false negative here. But we just don't know.
- ethbro 10y ago
  I wonder if there's enough play in the steering column that you could send generated micro torque amounts back up to the steering wheel, then see if they're resisted.
  - mikeash 10y ago
    You might be able to do it just by using road vibration that comes through the steering column. I think it's quite possible to detect hands with pretty good sensitivity in this way, it's just that the evidence suggests it's not actually that sensitive.
    - ethbro 10y ago
      Good idea. It's definitely an interesting problem and seems like it would be really relevant to the safety side of an autopilot. E.g. if there are no hands on the wheel, autopilot should probably do whatever it can to prevent a likely accident it sees

hashkb 10y ago

It might be difficult, but that shouldn't impact the right of drivers (if any) to try.

crispyambulance 10y ago
I agree, and it's not just drivers that would be able to make use of such "black box" log data but also insurance companies and law enforcement. The thing is it would require A LOT of cooperation from Tesla. Once Tesla starts sharing logs, they have to support it and can no longer change it at a "whim" for internal purposes. Perhaps most importantly, Tesla would also have to defend the accuracy of the logs when they go against other observations. Just because something is in a log, doesn't mean it correct, accurate or true. I can understand why they would exert a mighty effort to avoid giving out logs.

mikeash 10y ago

I'm skeptical of the "hands on the wheel" thing. They can't actually sense hands on the wheel, what they sense is torque on the steering wheel. I often have my car nag me to put my hands on the wheel when I already have them there, and I'm just not applying enough torque with them. It's possible that the threshold for the nag is just higher, and that the torque sensor is so sensitive there's no way it could possibly have a false negative here. But we just don't know.

ethbro 10y ago
I wonder if there's enough play in the steering column that you could send generated micro torque amounts back up to the steering wheel, then see if they're resisted.
- mikeash 10y ago
  You might be able to do it just by using road vibration that comes through the steering column. I think it's quite possible to detect hands with pretty good sensitivity in this way, it's just that the evidence suggests it's not actually that sensitive.
  - ethbro 10y ago
    Good idea. It's definitely an interesting problem and seems like it would be really relevant to the safety side of an autopilot. E.g. if there are no hands on the wheel, autopilot should probably do whatever it can to prevent a likely accident it sees