Now you are bringing in cost as another factor. I thought about mentioning that, but it was obvious and yet another tangent to the core discussion. I am quite willing to entertain the economic discussion about putting a price on human life (last I checked about $10M for the average american, which your hypothetical seems to support). However, this makes the problem almost impossibly complicated. In a world where Google Drive 4.2.3 is controlling all cars and the top speed is 137.339 we also save millions of hours of productivity wasted on the road and possibly a lot of fuel by not hanging out in traffic, etc. That alone may offset the extra cost of a self-driving car. So let's not bring that into the discussion.

The bottom line is that, all other things being equal, there is no logical reason for a machine driven car to have a lower fatality rate than a human driven car. If you want to write a law that allows variable safety laws based on the cost of the vehicle, I'd listen to it, but that is a slippery slope.

Care to tell me how it went when you tried explaining to one of your friends how self-driven cars can't evade potholes and won't be able to within the next 20 years? How long did it take for them to stop laughing? You're clearly grasping at straws if you are going to continue to defend such utter nonsense. This alone is enough to make your position ridiculous. (For further laughs tell them your fears of your self-driving car swerving into a tree because it was avoiding a squashed squirrel. You truly live in your own little bubble, don't you?)

BTW, there is a huge difference between accepting everything in this report and your claims that no self-driving cars will operate on regular roads in the foreseeable future. The latter WILL happen and it won't take 20 years. However, it might take 10, 20 or 50 years to have truly autonomous and passengerless vehicles on every significant road, but that will happen too.

Really? Why should the calculus differ whether a human is driving or a machine is driving? There is no logical reason for it, just emotional reasons, and do we want to kowtow to ignorant public opinion at the expense of human life (any more than we already do)?

Put another way, let's say that we are currently willing to accept 1 fatality for every 100M miles driven. If a self-driving car can drive 100mph with the same risk we should let it. Of course, knowing this new standard of safety is possible may mean that the public is now only willing to tolerate 0.1 fatalities per 100M miles which may limit self-driving cars to 65mph. BUT this also means that we should drastically lower the speed limit for human driven cars until they get to the same safety levels or ban them outright if it can't be done.

You might want to talk to one of those people who have followed AI for the last 50 years, they have some valuable insight to teach you. The initial promises of AI were indeed unrealistic which is why very few are tackling the problem of forcing computers to think like people, instead they focus on particular finite problems and solve those. And they've had amazing success, which you've seemed not to have noticed. Self-driving cars is one of these finite problems that you naively seem to think they're trying to solve by making cars think like humans.

Your pathetic example of a simple pothole shows how fearful you are of change and your general lack of knowledge and vision. Find the smartest person in the room you are in right now (someone, unlike me, who you respect and trust) and try to make the case to them that driverless cars can not now and will not in 20 years be able to detect and avoid potholes. Good luck with that. The reality is that not only can they do a decent job of it now (probably better than the average driver on an average day), they will continue to get better with every revision and they will do things that few or no human driver will be able to, e.g. they WILL be able to detect potholes that are hidden under water (unlike the superhumans you imagine are driving today) and they will be able to avoid those potholes without clipping the cyclist you didn't see next to you and without overreacting causing traffic nearby to freak out and they will be able to notify other cars that the pothole exists (especially if it's one of the rare cases that it couldn't detect it without driving over it) and it will be able to notify the city about the pothole.

You think you found a few examples today that driverless cars won't be able to handle? In the same amount of driving, the Google self-driving car detected 1000 possible threats that YOU didn't have a clue about. However, it's more humble than you are and doesn't rub your nose in it.

You can cling to your imaginary scenarios that you think self-driving cars will be unable to handle, there are actually some (albeit not the ones you mentioned). You can keep changing the definitions to make your predictions seem reasonable (your definition of "minor" road will become ever narrower, e.g.). This is one of those happy cases where you don't have to take my word for it, you'll see it slowly happening in the next few years (although your memory of this and your other conversations on the topic) will have been rewritten in your head.

Your fear is understandable and all too common which is why it will take longer than it should to get driverless cars on the road. But hey, a few hundred thousand lives is a small price to pay for keeping things the same as they ever was for as long as possible without scary robots taking our lives in their hands.

And I agreed with you, giving an example of how it would work with self driving cars "in the real world." However, this tangent was still totally unnecessary since the original post never stated that faster is always better - it was just an excuse for someone to flaunt their meagre knowledge of aerodynamics.

You clearly know little about self driving cars and perhaps technology in general. Just about every one of those situations the current best self driving cars could handle relatively gracefully, certainly better than a significant number of drivers. You seem to feel that it has to do better than the best possible driver (of which you are the one, apparently).

In at least one of the cases (possibly more) you are OVERreacting and causing a dangerous situation (while being the asshole you claim to think the auto might be). You absolutely should not stop for a random person standing by the side of the road because you think they might want to cross.

But hey, I have no doubt that you'll keep thinking that self driving cars will never be practical no matter what evidence is put before you, the same way you think machines won't ever be able to build products as well as humans, or that computers can't beat humans at problem solving in such as chess or say, Jeopardy!....

This is an excellent point, although I would state it more simply as "the top speed of a driving algorithm should be the speed at which it can drive as safely as a human driving the established speed limit." The established speed limit presumably reflects the amount of carnage we are willing to accept, which will no doubt change over time. Also, the top speed should be dependent on the car and its current condition - that is true whether it's autodriving or not, but will be easier to enforce in the future.

I especially love the idea that cars will be sold (or rented) based on these limits, which in fact spurs competition to make the cars safer. Most consumers could get their heads around that, whereas understanding the difference in 1 in 10,000,000 chance of fatal accident vs. 1 in 12,000,000 is incomprehensible.

Presumably the driver would have either direct control over such a preference (time vs. economy vs. safety) or the car would adjust to the route and or requested timing. But the post was merely referring to the speed limits and not the actual speeds so this is really an unnecessary tangent.

And yet it has inevitably happened time after time. What makes you think this will be any different?

Just as there are situations where human drivers are unable to cope. Except:

1) there are probably fewer that an ever increasing intelligent self-driving system would not be able to handle
2) these situations will diminish since these minor roads will evolve to be more autobot friendly (just as they once became more automobile (rather than horse) friendly
3) in those rare extreme conditions a remote driver can take over

As a practical matter, you're right, but that doesn't mean we should just roll over and accept it. The logical extensions of this policy are so far reaching that it would make our lives extremely constrained so we need to push back on this individually or collectively wherever possible.

Bingo. It's amazing that 200 other posters in this thread seem to think that any pilot, even a junior Korean pilot, would actually let 200 people due to his respect for authority.

That same day a bus crashed while driving from NYC to Boston and those passengers never reached their destination. In other words, your anecdote is meaningless - planes are still faster than buses.

You keep conflating the two interpretations of "more convenient". Is an apple more convenient than an orange? You could certainly make that point as you don't have to peel an apple to eat. An apple would be even more convenient if it didn't have a core that was hard to eat or had to be thrown out afterwards.

The list goes on, ..., ebooks can be bought instantly, often in places where the physical books are not available; ebooks can be stored on multiple devices, I can read the same book on the subway as I'm reading on my bedside; I can easily find the ebook rather than looking for which shelf or in which box I put it or which friend I loaned it to; I can travel with 50 different ebooks in my carryon rather than 1; I can instantly look up a definition, a translation, do a text search on the book, see all references to a character or place; I can easily read with one hand or lay it on a flat surface (for workouts, bath tub, ...); I can receive corrections and updates to the ebook automatically; and several more.

None of these features are "negated" by DRM. For you, personally, DRM is big issue, but it is clearly not for a huge number of people. All the problems you mention are trivial to me and many others and I expect you're exaggerating them for yourself as well. Let me address your issue with "Here is what an eBook is and this is what you can do with it." I primarily use Amazon, so I can easily tell you that it performs 99% of what I would do with a physical book right now and does a hell of a lot more. I have absolute confidence that I will still be able to read that eBook on any number of devices in a few months and I have little doubt that I will be able to access that book in 3, 5, 10 years. I even have a reasonable chance of being able to read that eBook in 20 or 50 years.

No, I don't know that every specific feature will remain the same, but I don't care (much) as long as the primary function still works, which it has a very high probability of doing in the near future and decent probability in long term. About as high (in the case of Amazon, at least) as it would for a physical book, actually.

In the rare case I want to "loan" a book to a friend I can send him a link electronically and he can download a copy of the eBook himself for a fee (which may very well be the same fee that I paid). All without the hassle of driving over to his house to hand the physical book, keeping track of it so I can pester him to give it back when he's done, only to find that he it's got coffee stains and many of the pages are wrinkled because he read it in the bath.

Paper books and eBooks are not the same, they don't have to do everything that the other does for them to be tremendously useful - this includes loaning or reselling. It would be great if eBooks also had those features, but that is a bonus, not a requirement. If you can figure out a way to get rid of the DRM while still protecting the publishers and booksellers, I suggest you let them know. They're not idiots and they're only as greedy as every other for-profit business - it's just that this is a difficult problem with no easy solution. The absolute freedom and guarantee you demand is not reasonable and you wouldn't want to pay the high cost to pay for it.

Well yes, you are exaggerating. A lot.

For every example you have that a printed book is more convenient there is another example where the eBook is more convenient. There is a use case for both, and some may prefer one over the other, but that does not warrant this one sided rant. If you want to argue that eBooks could be even more convenient, then be my guest - there is certainly room for improvement.