Nailed it.OP is an attorney and can comment, along with others... My understanding is that you should always first view the dashcam in private without law enforcement or anyone from the other party viewing. The camera footage is your own property and if it shows something incriminating, you do not have to produce it to others. By viewing it for the first time with police, you give up that ability.
Umm.....Read what you wrote again. If the thing is an AI, you don't program it in any traditional sense.Exactly! This is the kind of data that the AI has to account for in FSD. Random, human, actions that defy logic but happen all the time. How do you program that avoidance behavior into CyberTruck's FSD?
As I understand Elon's "video learning" for FSD development, video captured by the Tesla fleet is uploaded to a Super Computer that plays the video to an AI that learns from the videos how to drive. The AI can then upload to the Tesla fleet improved FSD algorithms for autonomous driving in real time through the neural net. I know it is far more complicated than that, but, it generally describes the Tesla FSD autonomous solution. I also have those drones but the AI choices for flight path are human written code based on obstacles its cameras see.Umm.....Read what you wrote again. If the thing is an AI, you don't program it in any traditional sense.
I have a small consumer-grade drone that actively avoids obstacles occurring anywhere - up, down, sideways, doesn't matter. It routes around them if possible to continue its flight path.
It is disappointing. However, I can still drive and enjoy it until it goes in shop. I'm lucky it wasn't damaged seriously.I'm Cyberbummed, what a Cybershame. This is fucked up. Now you have to stop enjoying the CT, and concentrate on fixing a truck no one knows how to fix yet. ?
Cyber
That is true. But, if you activated "DashCam" and I can prove you did, yet, the footage is deleted. I can ask for a "spoilation of evidence" charge which allows the fact finder to consider its disappearance evidence of your negligence. Best not to activate it at all if you want plausible denial in effect.OP is an attorney and can comment, along with others... My understanding is that you should always first view the dashcam in private without law enforcement or anyone from the other party viewing. The camera footage is your own property and if it shows something incriminating, you do not have to produce it to others. By viewing it for the first time with police, you give up that ability.
I agree with that, and "training a model" is one way to describe the process. But training is generalizable, not specific.As I understand Elon's "video learning" for FSD development, video captured by the Tesla fleet is uploaded to a Super Computer that plays the video to an AI that learns from the videos how to drive. The AI can then upload to the Tesla fleet improved FSD algorithms for autonomous driving in real time through the neural net. I know it is far more complicated than that, but, it generally describes the Tesla FSD autonomous solution. I also have those drones but the AI choices for flight path are human written code based on obstacles its cameras see.
You would have to prove beyond reasonable doubt that the footage was deleted with the explicit purpose of hiding evidence though.That is true. But, if you activated "DashCam" and I can prove you did, yet, the footage is deleted. I can adk for a "spoilation of evidence" charge which allows the fact finder to consider its disappearance evidence of your negligence. Best not to activate it at all if you want plausible denial in effect.
Oh I appreciate your thoughts about how it "trains" or will apply it's learning to reacting in real world situations. It is a cool thing to speculate about as the FSD tries to reach Level 5 autonomy. My situation is a great example of "human choice" to make a "U-Turn" at an intersection where it is prohibited. How will the FSD react in real-time to that scenerio? What will it tell the CyberTruck to do?I agree with that, and "training a model" is one way to describe the process. But training is generalizable, not specific.
So I am noting that specifically having training examples of "million to one or billion to one" circumstances to train in order for training to happen that will affect the machine's response to those circumstances. With LLM AI models (which the Tesla is not) the ability to generalize is a measured performance dimension and is highly significant.
I doubt if the Tesla computer cares if a car coming at you from the side is from the side or the back or the front. It's a threat and it should react to that.
Of course I am just guessing here.
You can ask in interrogatories, questions at depositions, or subpoena the information from Tesla. The driver can lie of course, and/or Tesla can refuse to provide the video captured based on privacy arguments. But, if the wreck is bad enough with injuries severe enough, a determined personal injury lawyer will find out.No one can prove anyone has dashcam video unless they see it first or it’s shared openly.
Every jurisdiction's laws are different. I can only speak about Alabama. I will tell you, a friend of mine recently defended a man accused of vehicular homocide in the death of his girlfriend and unborn child. The issue was whether he was criminally negligent in the operation of his Hyundai car at a high rate of speed in the rain. On his way to church, he lost control of the car and crashed killing his girlfriend and she was 7 months pregnant. The *black box* said he was going 93 mph. He testified and a witness testified he was driving 60-65 mph. The State's case against him was all built upon the digital record of his speed and that was obtained by law enforcement without his permission.You would have to prove beyond reasonable doubt that the footage was deleted with the explicit purpose of hiding evidence though.
"Your honor, I was so frustrated and worked up I hit the wrong button apparently".
Not going to be easy to prove that in court I think lol