It sounds like you're trying to say that there's no such thing as automation, which is kind of silly (especially when you give a decent example). Listing occupations for you would be fool's errand, because no matter what choice I make you can take your "just assisting" argument to a ridiculous extreme, either by saying that there need to be people to make, program, maintain etc. the machines or that those jobs have not gone away because someone somewhere in the world is still doing them. Traffic signaling police haven't been completely replaced by traffic lights because someone needs to set the timing or change the bulbs. Movable type printing typesetters haven't been completely replaced because authors still need to type into a computer, etc. That's absurd.

Actually, I just have less faith in human drivers than you do. All your examples show that humans are horrible at this kind of thing and commonly make mistakes that no production level self-driving car would even get close to. Your slipstream scenario is a perfect example of why we need to move to self-driving cars as quickly as we can.

I disagree about the progress, but that is just semantics. The first 90% of the problem was relatively straightforward, but you can't expect the same kind of progress in the last 10% (or it would have been solved long ago). If we got to 90% in the first 50 years, getting to 95% in the last 15 may not seem like progress, but I think if you looked at a passage translated with 1999 technology vs. the best available today you would find a marked difference. Even freely available services like Google do a very good job at translating between western languages, although it will be much longer until they can handle Japanese or Tamarian.

I don't know why you insist on talking about absolutes, e.g. when no human assistance will be required for any particular activity. Machines have been replacing manual labor, and replacing human translators and are starting to replace transportation workers. For the first two we are well on the way (90% or more) and we will likely never get to the point of replacing humans 100% in any of those areas. It's just impossible to argue that the machines are getting better at these jobs faster than we are.

You can become skilled at anything, including nail guns. That is not the point, I'm not saying that most carpenters aren't skilled or that some guys can't wield a nail gun like a maniac - they can. But it doesn't take much skill or training to pick up a nail gun, point it at a wall and pull the trigger. I'll agree that the ability to use a nail gun is a skill, if you agree that frying an egg is a skill, vacuuming a house is a skill, and picking up a box and moving from one place to another is a skill.

I'm not sure what your point is. It's obviously true that machines and computers have a harder time at some kinds of problems than others. It's also true that they can't do everything a human can do today. Nobody would dispute that. However, everything talked about here, nails, cars, translation are more automated than than they were 100 years ago and various technology improvements and cost efficiencies will keep that trend going.

Specifically, nail guns are just one step on the road to automating construction, although not even the first step. As I alluded to, many things that were assembled with hammers and nails 200 years ago are now done by machine, even many parts of building and home construction (e.g. window frames). When I said there was "not much skill", I was specifically referring to the line "hold the hammer properly and swing it". But I'll go further in that for most construction where to put the nails is not a skill of the guy holding the nail gun, it should be determined by the architect and the building codes. I don't consider myself a skilled carpenter because I can put together an IKEA bookshelf.

There's absolutely no reason why a carpenter bot couldn't slam 24 nails in a 2x4 AND verify that they were done correctly, it's just not cost effective to do, for now. We are not going to completely get rid of carpenters any time soon, but their jobs will slowly go away with more automation. In fifty years you might have a house built by a single foreman who's job is primarily just to oversee the bots.

As for your translation example, yes translation is hard, but the bulk of translation done today is completely automated. It's not terribly good, but it is good enough for most uses; Facebook, website browsing, manuals on cheap chinese electronics, etc.

For nails, words and cars, the ASSISTED part will become smaller and smaller.

If violating the traffic laws doesn't cause an increase in risk then they shouldn't be laws. Which laws are you proposing to get rid of, or is it all of them?

It's a lost cause using logic and math to convince some people that automated cars will be safer. They will always come up with one obscure scenario that computers just can't handle, but this is the first time I've heard of one with a duck and unbelted kid - that's original. However, if it turns out that there's a rash of duck/kid accidents it would be pretty easy for Google to require all passengers be belted while the car is in motion.

That's a terrible analogy. The vast majority of times a metal cylinder is inserted into another object it is done by machines run by computers. Plus, the majority of times a human makes a conscious decision to put a nail into an object, it's done with a nail gun, no swinging, not much skill, the machine doing almost all the work. Only a relatively insignificant percentage are done by a human picking up a nail and hitting it with a hammer, and those cases are becoming fewer all the time. It's not magic.

I have to say that I think your scenario is far fetched to the point of being absurd.

The very FIRST thing an automated car (or the programmer) would do would be to NOT put yourself in risky situations. IF the rolling box situation is common enough to worry about then the car would never get into the slipstream to begin with. Your entire hypothesis is flawed: you assume the car will slipstream because it can do so safely, yet then you go on to demonstrate it's NOT safe.

Ignoring that contradiction, the automated car can still do any number of things when it gets into that situation to minimize danger. IF they decided to put in special code allowing the car to slipstream they would certainly have a number of parameters for when it is safe, as AC pointed out. If the risk is too high when a vehicle is behind you, the car would back off and get out of the slipstream. If it were boxed in, it would see even higher risk and be sure to back off. This is not that hard. In fact, you don't need special "slipstream" code to do this, an automated car should ALWAYS increase the distance to the car in front of it when it is being tailgated and it will ALWAYS avoid being boxed in (slowing down so that it has an empty area next to it).

Even ignoring all of that, the automated car will handle the rare, rare situation of having an accident in the "rolling box" safely the vast majority of the time, it might be alerted to the accident up ahead well in advance and it will detect it on it's own far sooner than a human would. It will react to what all the trucks around it are doing and pick the safest option.

However, let's assume that in this rare, rare case the car logic goes haywire and explodes the car incinerating all passengers. The number of lives saved by all the common accidents that are avoided are well worth the occasional spontaneous combustion (especially when it's only people who knowingly take that risk to save a few bucks on fuel).

Driving is inherently dangerous, every time you are on the road, even following every traffic regulation to the letter you are putting lives at risk. When you violate those laws you become even more of a risk (in the vast majority of cases). I will fully grant that you are a far better driver than the majority of drivers (and probably much, much safer than the average cabbie), but that merely means that you are adding less risk, not no risk. You may be shooting a gun off in a remote area, but you're still doing it. You absolutely cannot guarantee me that you will never get into an accident, that's just crazy talk.

I don't know if you're aware of the concept of external costs, but while the things you mention here are a benefit to you and your passenger, they have costs for everyone else, like increased risk or delaying other people on the bus/train. It MAY be that you get enough benefit to outweigh their costs, but you have to grant them the same option. Which means that they also get to speed and/or make you wait for the train AND a self-driving car can make those decisions more fairly and consistently (eventually).

I don't know about the car in this particular video, but some versions can already detect potholes and navigate safely around them. But they also will be able to detect them in the dark or when they are filled with rain water, and they will be able to record and notify other cars so they can avoid them and the city so they can fix them.

Google, Yelp, Chowhound, ... there are already plenty of services that have more knowledge about local restaurants and bars than you do and they are constantly adding more data and getting better at making recommendations. The panel on the self driving taxi will not only find a good restaurant you like it will tell you exactly what time they close, it will make reservations for you, etc.

It may be several years off, but the self driving taxi will also be able to tell the bus to wait for you and if it were me who designed the system I would put in a cost for doing that.

The average NYC cab driver, for example, provides a terrible experience and few if any of the services that you offer. It wouldn't take much for an automated taxi to outdo them in safety and service. Not to mention that it is better for society not having 10s of thousands of taxis driving around looking for fares - automated taxis wouldn't do that.

I have used car services and chauffeurs who offer the same premium service that you advertise. They are much better than the average cabbie in almost every way and, of course, they charge for that. This merely buys you some extra time. Some people will continue to pay extra, including those who distrust technology or just want human contact. Of course, all the other displaced taxi drivers will want to move into your business and at least some of them will be qualified and compete for your premium business.

In the near future I expect a large number of taxis, if not all, to be a hybrid - a self-driving car that includes a driver behind the wheel. This is a decent compromise for everyone until the technology becomes bulletproof. This might also be a job in your distant future, you still get to load bags and chat with the customers who want that, but don't do the actual driving.

If you are significantly better than the average server then you should command a higher salary from your boss.

The server is the person most responsible for the overall quality and enjoyment of your meal and typically the only person you're going to be interacting with. They, in turn, tip out to the support staff so that they (the server) get what they need (water refilled, drinks poured, food in a timely manner with special orders considered, ...). It's not the only way to do it, but it is rational, for the most part.

If people don't get hurt by you violating the traffic laws then that is just luck. It's like shooting a gun into the sky, not a sin, but don't pretend it's 100% safe either.

The point is that you want us to believe you are providing a necessary service by breaking the law and endangering (even if only barely) the lives of others. This may very well be a differentiator for you when competing against self-driving cars that are required to follow the law, but you can't expect society to approve it. If we did then we could just as easily turn off the limits on self-driving cars and let them speed, roll through stop signs etc. and they will be able to do it more safely and more efficiently than you can.

If you are retiring in the next 10 years then you probably have nothing to worry about, but if you think you can continue to earn the same money doing this job in 20 to 30 years I think you'll be unpleasantly surprised.

It's not that they don't know how to do this kind of thing, it's just that they can't do it perfectly yet. They are not going to claim such features, let alone let them out in the wild until they can do much, much better than the average driver. Self-parking technology, for example, already exists on the streets and works pretty well, but still requires at least a minimum of driver interaction (e.g. you have to tell it where to park and line it up). But even with current technology there are plenty of workarounds that will extend the usefulness of these cars, e.g. when it takes you to any number of known lots it doesn't even have to navigate a complex garage to find an open spot, it can just drop you and then go wait in a designated area with all the other driverless cars (just as chauffeurs commonly do).

I suggest you have a backup plan.

If we, as a society, are okay with you (taxi drivers) speeding up to shave off a few minutes, then we certainly should also let a self-driving car with a better safety record do so as well. This is not a point in your advantage.