Autonomous Vehicle Maker Consolidation, Bankruptcies and Negligence Lawsuits are Coming
The hype in the autonomous vehicle industry is beginning to wear off. Those bogus miles driven and disengagement metrics are starting to look awfully suspect in light of the significant lack of progress we are now seeing in the industry. A major sign of that being in recent articles on Waymo. The Independent did a great job of documenting their difficulty in turning left, navigating intersections and on ramps as well as those driving around them being frustrated by their unnatural actions. What is the point of advertising the miles you have driven or how your disengagement count is going down if you can’t navigate benign scenarios in well-lit and marked areas in good weather? (Chris Urmson one of the founders of this industry is the ONLY voice among the AV makers I have seen to affirm this.)
There are two reasons we will see companies starting to go bankrupt and experiencing an increased level of lawsuits.
The first is that it is not possible to create an AV anywhere near Level 4 using public shadow driving to develop them. (That is the process where the driver allows the vehicle to drive to test how well it has learned. Then taking over when it makes errors. I am NOT referring to using data collected while humans drive the vehicles. That is encouraged.) It is impossible to drive and redrive, stumble and restumble on enough scenarios to get anywhere near finishing. Especially complex and accident scenarios. (RAND said it is 500 billion miles. Toyota says it’s one trillion.) Uber spends $200M a quarter trying and are no further than the same first base Waymo is on. Chris Urmson himself has evolved his thinking along these lines and has said it could take 30 years to get this done. (Since Aurora shadow drives as well it’s clear he doesn’t quite get it yet himself. I do however give him a lot of credit for letting objective data drive his thought process and having the courage to say so in public. He has also said handover/L3 should be skipped. As has Waymo, Toyota and Volvo. Problem is ALL of them still use the practice to develop their technology. Waymo actually said and started off with not having shadow drivers in vehicles with the public in them. They have however pulled back on that policy. Tesla, Comma.ai and PolySync go further and use their customers as the development Guinea pigs.)
The other issue is safety. Causing injuries and casualties for no reason. The first part of that revolves around accident scenarios. What do you think the public, law enforcement, attorney, the press and government officials will do when they realize the only way to learn how to handle the thousands of critical accident scenarios is to experience them over, and over and over? What happens when that first child or family dies needlessly? The other issue is handover. The act of allowing the vehicle to drive until it can’t handle something and requires the driver to take over. Be that in development of the technology through public shadow driving or in actual public use as an L3 system. No matter what monitoring and notification system you use there is a period of time that no human will have enough time to regain enough situational awareness to do the right thing the right way. Virtually every accident scenario has this problem.
Right now the AV makers are scared to death of causing that next Elaine Herzberg casualty. Or god forbid a child or family perishes. At some point though investors are gonna figure out that miles driven and less disengagements mean nothing without knowing what the scenarios are. Unless these companies push forward to more complex and dangerous scenarios soon all they will do is raise the redriven mile count and hype and burn through investors money making zero progress. Odds are those investors don’t stand for that and force the AV makers burning through their money to move toward those complex and accident scenarios. And that’s where the lawsuits will come. That in turn will hasten those bankruptcies and fruitless consolidations. Either companies go bankrupt because they are burning cash in a constant loop of running duplicate scenarios making no progress or they burn through cash settling lawsuits. The actual result will be some combination of both.
There is a solution. That is skip level 3 and move to 99% simulation use for development. Why isn’t anyone doing that? Because the simulation technology this industry understands and is aware of, built mostly from gaming technology and automakers is not adequate. In fact, it often provides misleading results. The remedy there is to create an end-state scenario matrix integrated with aerospace/DoD simulation technology and FAA-like validation processes. (This is where folks normally say that air travel is nowhere near as complex as the AV world. That is true. What is as complex and analogous are DoD networked simulated urban war games.)
Please find more one this here
Impediments to Creating an Autonomous Vehicle
The Dangers of Inferior Simulation for Autonomous Vehicles
Elon Musk is the Pied Piper of Autonomous Vehicles
Uber is making a fatal mistake
Waymo is the Wolf in Sheep’s Clothing — On Lower Ground than Tesla
