Tesla is exposing Autonomous Vehicle Industry’s avoidably Dangerous and Impossible Engineering Approach
Elon Musk upped the anti this week by stating Tesla’s Autopilot (AP) will reach L4 this year. And eliminate the need for a steering wheel next year by reaching L5. In the same week Chris Urmson upped his estimate for the industry to get to L4 to 50 years. (Several years ago, he said it would be this year.) Ford showed both ends of the spectrum by first saying they under estimated the task then saying, after Musk’s statement, they too will reach L4 is some areas next year.
What Elon Musk has done is shine a solar flare on the industry. In order to make his reckless and not remotely possible promise, he will have to either stay on his current path and blame someone else for not making it, lie about the system capabilities, which will be easily proven wrong, or he has to try and get there. Given his track record he will probably do the latter. This will mean accelerating the training of his systems on complex, dangerous and actual accident scenarios. That in turn will greatly accelerate the death rate of his overly trusting human Guinea pigs, their families and the people around them. Think about this. They have to use the same process to train their systems on deadly scenarios they use for the benign ones. That means experiencing them thousands of times over. Given Tesla’s extreme misunderstand of what proper simulation can do (aerospace/DoD based technology not the ever-present gaming-based technology in this industry) this means most of it will be done outside of simulation. This will lead to thousands of deaths and virtually guaranty, quite unfortunate, the first child or family is needlessly killed this year. What do you think will happen when the public figures this out? When the press and the government figures this out?
Some other companies in the industry including the AV industry’s PR firm PAVE Campaign have (finally) spoken out against the time line and the use of the term “Autopilot” and others that mislead people into thinking these systems are fully driverless. But what they have not said and cannot say because they DO THE EXACT SAME THING is speak out against the approach of using public shadow and safety driving for development and testing. This means they too will have to take the same path as Tesla. Meaning they will kill people needlessly as well. Only they will spread those tragedies out over a longer time period. The worst part of course is that conventional wisdom is wrong. This is not the best or only process to develop these systems. And the deaths it causes are not for the greater good. The truth is the process can never lead close to L4. This means the lives that would save will not be saved. Worse of all lives have been and will be taken needlessly as they perpetually fail. It is impossible to drive the one trillion miles or spend over $300B to stumble and restumble on all the scenarios necessary to complete the effort. Many of which are accident scenarios no one will want you to run once let alone thousands of times. Also, handover cannot be made safe for most complex scenarios, by any monitoring and notification system, because they cannot provide the time to regain proper situational awareness and do the right thing the right way.
The solution is to replace 99.9% of that public shadow and safety driving with proper simulation. That being systems based on aerospace/DoD simulation technology. To make matters worse Musk and his AI lead Andrej Karpathy made more inaccurate statements this week concerning their belief that the real-world has to be used for most of this. That of course is ridiculous and covered in my first article below. I also cover why most of the simulation being created or sold in this industry is inadequate, will cause false confidence and lead to real world tragedies and debilitating rework.
Using the Real World is better than Proper Simulation for Autonomous Vehicle Development — NONSENSE
The Autonomous Vehicle Podcast — Featured Guest — —
SAE Autonomous Vehicle Engineering Magazine-End Public Shadow Driving
Common Misconceptions about Aerospace/DoD/FAA Simulation for Autonomous Vehicles
The Hype of Geofencing for Autonomous Vehicles
How Driverless Vehicle Makers Should Prove their Technology Works