They are not a pipe dream in Silicon Valley, and may not be a pipe dream on dedicated highways that only allow automated cars.
 
You are mostly correct, any Google car that lacks manual controls will be grounded during bad weather and/or novel conditions since 'autonomous' parts heavily relies on detailed mapped and predictable environment.

Traditional car makers (e.g. Detroit 3) are not always wrong and in this case Google should not be simply assumed to be correct. Since I was not part of these meetings, I can only form my opinions based on what was reported. Still, there are some things that concern me with Google/Tesla approach to autos:

* Unwillingness to finalize the product is part of Silicon culture. When I buy a car, I expect final product with very rare instance of patching (e.g. recalls) and no instances of altered or added functionality. The fact that when you buy Tesla you are subjected to "patch Tuesday" tinkering greatly worries me.
* No defined model years. With traditional cars you usually know that parts from years X-Y models A-Z are interchangeable. Not so much with Tesla - where mid-model changes are commonplace. What going to happen when 10+ year old Tesla needs a new part? Always buy new, because no two of them are ever the same?
* Used car market. For electric cars it doesn't exists. This means that depreciation on these is largely unknown.

To science denialists above, I hope you apply equal skepticism to medicine and health and food safety aspects of scientific knowledge. This way your views will have self-correcting, instead of humanity-correcting impact.

CO2, a greenhouse gas, is tied to climate change but is also expected and always-present part of our atmosphere. Why do we need satellite to know this?

Can we have a better summary?

No, "most sane people" will not comply, because everyone is 'above average skill' drivers, and car wrecks happen to 'other people'.

Don't underestimate addictiveness of "always connected" lifestyle and power to rationalize away your bad decisions.

Libertarian market driven approaches of 'perfectly informed' customers having access to 'flexible supply' are only workable on paper. Sure, it would be nice if we could get there, but meanwhile our situation continuing to deteriorate. Time to abandon this quixotic quest.
 
What we need is "mostly works for most people most of the time", and to get there we need policy with teeth that mandates Net Neutrality. Sure, it won't prevent all abuses, but we only need to prevent worst of them and let the rest play out in courts.

Without FIPS certification system engineers won't be able to include BoringSSL in US-government facing applications, since doing so will disqualify them from procurement lists. Since US gov't is largest consumer of cryptographic products in the North American market, BoringSSL must certify or stay irrelevant.

Make sure to run it on DVL distro

Very interesting post, thank you for writing it up.
 
I have a question. Are there guard bands in biological computation (e.g. our brains) ? I was under impression that our cognitive processes (software) are optimized for speed and designed to work with massively parallel but highly unreliable neural hardware.
 
What I am trying to say is that nature performed optimization decided that it is better to be very efficient all the time, and correct some of the time, but also be very good at error checking. While our CPU and OS designers decided that computing devices must be correct all the time, efficient some of the time, and poor at error checking.

Could someone explain to me why further refinement of fabrication process is the only way to progress? With a car analogy?

If you are attempting being rational, then you need to reevaluate your decision making process. With rational approach, cost of ownership paired to a list of features is all you should consider. Optimizing the curve, you eliminate initial depreciation by buying used, and high repair/maintenance by selling the car before it gets too expensive to maintain. The optimum ends up being 2-8 years old car of any reliable car. Why reliable? First, it impacts resale value at 8 years old. Second, it allows it to make it to 8 years old without incurring major repair costs. You can't get any newer than 2 years, there isn't reliable supply of these, and selling before 8 years old leaves you with too short of an ownership to flatten the depreciation curve.

In closing, you might not want to drive that hypothetical lasts-50-years car in 49th year, but it makes cost of year 1 to year 5 cheaper.

Again, I disagree.

Very few people outright buy the car and use it until junk yard. For people that lease resale value directly impacts their payment, less it worth at the end of leas term higher the monthly payment. For people that finance, trade-in value of their last car impacts their payments. In almost all cases original buyer and second-hand buyer are financially tied via 'cost of ownership' concept.

The car is not a software product, and obsolescence is not clear-cut or binary. Nearly all car features, including safety, are tied to original car budget. For example, luxury 5 year old car will likely still have more safety features than 0 year old econobox. That is, you are A LOT more likely to survive a crash in 2009 Mercedes S500 than 2014 Chevy Spark. I can find many examples of 10 year old cars that are still 'feature-complete' with anything that can be found on the road today. Your "obsolete after 5 years" is way, way off and suggests to me you are trying to rationalize your leasing habit.

What is the issue and why is it not covered under 10 year / 100K warranty?

If you ever visit automotive junk yard you will see that write-off accidents are minority and mechanical failure is majority of causes that lead cars to end up there.

Mechanical failure could be further categorized to catastrophic failure (e.g. timing belt in interference engine) and multiple minor concurrent issues that exceed replacement value. While I don't have hard data on this, I believe that leading cause of why cars end up in junk yards is transmission failure. Still, most of the cars that make to the junk yard could be made run or are still running.

Therefore, it is safe to conclude that car with higher degree of mechanical reliability, and cars that are cheaper to repair will be used for longer time.

>>> A car that falls apart after 5 years isn't any higher quality than a car that runs for 50 years, if you're going to replace either in 5 years anyway.
 
Faulty thinking. While you might get tired and replace car in 5 years, a car that runs for 50 years will have multiple owners. Its residual value will be higher. Environmental impact of manufacturing and then recycling it will be lessened due to getting spread over many more years.
 
  Car that runs for 50 years is always higher quality that on that falls apart after 5 years no matter how you use it.