Just to play devils advocate here. Assuming Uber gets a permit to test autonomous vehicles in public, how much risk has been mitigated as a result vs not having a permit and allowing them to continue testing?
Reading the legislation, the only requirement for the permit are:
- have insurance with up to $5 million
- the driver has a drivers license and has "Instruction on the automated driving system technology", as well as defensive driving training
Those both sound like reasonable, common-sense things to have. I'd also want a company testing their vehicle to have insurance and know how to drive their car. But at the same time if this permit didn't exist, how much less safe would the world be? Really?
They're not stress testing the software, they're not analyzing the code, they aren't testing the competency of the people who built the software, etc, etc. They aren't doing any of that for a good reason. So ultimately this is mostly just theatre or an exercise in letting Uber know who the boss is.
Because I highly doubt Uber doesn't have the insurance or experienced drivers. Nor do I think there will be any significant amount of tech companies wrecking havoc on California because they didn't think to buy auto insurance before starting an automobile software company, save a permit.
CA DMV also requires companies testing autonomous vehicles to report all their crashes and disconnects. Here are all those reports.[1][2] 20 companies have signed up. This is basic data collection to help decide when a system is safe for deployment.
Uber apparently tried to evade this minimal level of scrutiny. Looks like it didn't work.
I can think of a number of ways to collect this data without requiring a special permit. The police are in coordination with California, so incidences involving semi autonomous vehicles can be collected using existing systems. I'm also sure there are means to shutdown Uber's autonomous testing via court order.
Additionally insurance and drivers licenses are already regulated.
I'm still not convinced of the necessity. The hammer is always looking for a new nail.
The bar for expanding bureaucracy and adding highly specific oversight is so low and so commonplace that you're labeled a techno-libertarian for questioning the basic economic cost (realistic benefits vs tradeoffs) of doing so.
You may feel more comfortable knowing someone is watching over these specific companies testing process. That doesn't automatically mean the world has been made safer as a result of their time spent writing complex legislation, building regulatory systems, enforcing it, slowing down development of technology, battling cases in court, pushing companies to other states/countries, etc.
Unfortunately questioning the realistic utility of such regulation is no longer the default and is dismissed as the foolishness of some silicon valley tech people disconnected from real life...
The data collection required for autonomous vehicles includes even very minor accidents. The idea is to collect data on what failed and why before somebody gets killed. Unlike regular accident data, all reports are public. This lets us see that Google's repeat problem is being rear-ended when their cars stop while entering an intersection because they detected cross-traffic. It also tells us that Cruise Automation hit a parked car on 4th St in SF for no good reason.
Anybody testing autonomous vehicles has this data. They just have to send it in.
So do you think the public wouldn't be able to inform themselves on which service or vehicle is safer before purchasing these cars without this data? Or that Google, Uber, Tesla, etc wouldn't already be taking these numbers very seriously and doing everything to minimize them? Or that the public wouldn't truly know the real safety of these vehicle as a whole vs human drivers.... without this mandatory test vehicle permit?
> Or that Google, Uber, Tesla, etc wouldn't already be taking these numbers very seriously and doing everything to minimize them?
Given that we know existing car manufacturers have faked or covered up safety data in the past, you're putting a lot of faith in the goodness of tech companies.
The organization who detected that VW was faking the data wasn't actually the government (California Air Resources Board in this case). It was detected by a privately funded organization... funded in part by technology billionaires:
...I'm the one questioning the utility in of California's system of control. My faith in either actor's earnest participation is therefore not of much use.
I do know customers value safety and honesty of companies they buy/use cars from, and so do the investors who back these companies. Which are the people who keep these companies in business. That pressure exists independently of these permits. Which is why one must realistically question the utility of special government oversight.
I'm definitely a fan of the thesis that government should be more competent, although I don't think that's where you were going with that. :) I'd love to see government regulations that demand that you should stress-test the software, hire someone competent to analyze the code, get licensed engineers to be involved in design and development-process review, etc.
The difference between your devil's-advocate worldview and my worldview is I think not in our perception of facts - we both agree that the government isn't super competent right now - but in our optimism. You (or your viewpoint) isn't optimistic that government regulation will ever get good, and would rather just stop trying, but is pretty confident that individual humans and companies are unlikely to be abusive, or to cause much damage if they are. I'm resigned to expecting that someone, somewhere, will try to get away with the bare minimum required, and in the absence of regulation will just be too stupid to realize they're risking people's lives until people die. But I do have higher expectations of government regulation, and believe that it can be good or at least we can work towards it being good.
(Also, government regulation doesn't have to involve the government doing things. Several industries have so-called "self-regulatory organizations", which mostly exist because of the implicit threat that the government could start regulating more heavily. That is, the SROs only exist because of government letting industry know who the boss is, so there's a distinct positive effect of that posture! The SROs aren't perfect, but to first order everyone is better off: the government isn't interfering with things they don't understand, yet actual, competent oversight with meaningful enforcement powers exists. Self-driving cars are too small an industry at present to meaningfully self-regulate, but it certainly sounds like they won't be very shortly.)
Reading the legislation, the only requirement for the permit are:
- have insurance with up to $5 million
- the driver has a drivers license and has "Instruction on the automated driving system technology", as well as defensive driving training
Those both sound like reasonable, common-sense things to have. I'd also want a company testing their vehicle to have insurance and know how to drive their car. But at the same time if this permit didn't exist, how much less safe would the world be? Really?
They're not stress testing the software, they're not analyzing the code, they aren't testing the competency of the people who built the software, etc, etc. They aren't doing any of that for a good reason. So ultimately this is mostly just theatre or an exercise in letting Uber know who the boss is.
Because I highly doubt Uber doesn't have the insurance or experienced drivers. Nor do I think there will be any significant amount of tech companies wrecking havoc on California because they didn't think to buy auto insurance before starting an automobile software company, save a permit.