The premise of the article Sex distracted Driving on the rise is that more and more people will use the spare time and energy of level 3 and 4 self-driving cars to perform sexual activities. Whether this is frequent enough to measurably increase the death or accident rate I think is highly suspect however it presumes that self-driving is going to be prominent soon and there is a part of me that seriously questions aspects of this.
One can hardly find more excited and optimistic person about technology and Elon Musk’s ventures and vision than me. I have been writing about the future of technology for a long time and building that future but I know there are also limits.
First I will talk about self-driving cars and the problems and path to the future. Then I will talk about Hyperloop. Hyperloop is important as you will see because I believe it is inevitable that not only will hyperloop happen but we may want to build an aggressive plan to make hyperloop happen sooner because it will lead to the point where transportation costs drop to nearly zero variable cost.
The combination of these 2 technologies will have a profound effect on our daily lives similar to the smartphone or the internet.
Self Driving Cars
I will talk about the short term problems that will make this technology seem to take forever to get to where it seems common and reliable. I will then describe a world where a lot of our problems related to transportation goes away. Ironically as self-driving cars become common in 20 years a new technology will start to gain traction that will make the car a far less useful vehicle.
The US government defines self-driving this way:
- Level 0: The driver completely controls the vehicle at all times.
- Level 1: Individual vehicle controls are automated, such as electronic stability control or automatic braking.
- Level 2: At least two controls can be automated in unison, such as adaptive cruise control in combination with lane keeping. Example: Tesla Model SLevel 3: The driver can fully cede control of all safety-critical functions in certain conditions. The car senses when conditions require the driver to retake control and provides a “sufficiently comfortable transition time” for the driver to do so.
- Level 4: The vehicle performs all safety-critical functions for the entire trip, with the driver not expected to control the vehicle at any time. As this vehicle would control all functions from start to stop, including all parking functions, it could include unoccupied cars.
Tesla and other manufacturers have demonstrated Level 2 pretty well. Some will call the Tesla a poor level 3 system. Google makes a prototype car which has effectively done level 4 on city streets at slow speed.
What we have done so far
Google has demonstrated that for cars driving 25mph or less on the streets a self-driving vehicle can operate as safe or safer than humans and can navigate the well known streets of Mountain View, CA.
I personally own a Tesla which I have reviewed with glowing praise twice. I have experienced the self-driving features. The Tesla is very good at maintaining lane on a normal highway and speeding up and slowing down when required. This is incredibly useful especially in high traffic times and it is effectively a second pair of eyes that I feel enhances my safety enormously. I applaud the efforts and look forward to improvements but I can also see how far away we are.
Tesla is not as far along as Google in some aspects but I don’t believe Google has solved some of the critical issues to roll out this technology as level 4 and even unsure if level 3 is possible realistic for many driving situations.
It is important to note that several really intractable problems are in the way of a complete driverless experience to become uniform and ubiquitous. If you hear that it is just a matter of rolling this technology out to people you would be hearing a lie.
The problems with driverless cars software and hardware
The basic first problem is reaction time.
In a level 3 system there is a need for the driver to take over in a dangerous situation. Basically, when the Tesla loses visual ability to see the lane markings or finds cars too close it gives up and it really has given you no/very little time to react.
The typical scenario I have seen for this is when at a complicated intersection of cars on the highway with exits. When it loses the lane you are in a tremendously complicated situation with cars all around you going in and out of lanes exiting and entering at fairly high speed. This is hard enough if you had been paying very close attention. To do so in a fraction of a second is extremely dangerous.
I grant that the Tesla software is “alpha or beta” and they are expecting their loyal drivers to be cautious and smart. Indeed they are in general. The problem is both simple and complex. By the time any computer recognizes the situation is out of its control (level 3 systems) there is no possibility for the human to react in time to “take over.” Level 3 systems have the flaw that by the time the computer recognizes it can’t handle the situation there is no way you can figure out what’s going on and fix it.
Google has avoided this by mainly operating on city streets at low speed allowing enough time for the car and other drivers to react and prevent a disaster even in some cases if other drivers turn out to be malevolent.
AI required sometimes
I have noticed that there is a certain amount of “AI” that all humans do when driving. We use experience to estimate the lane when there are no markings, we use experience and some innate sense of how other cars move or will move relative to us. We expect cars around us will act rationally so we give tolerance far beyond what could ensue if people acted bizarre. We use lots of cues to tell us things that a computer would have to learn and more important may become extremely complicated to program as straight physics and logic.
An autonomous vehicle has many advantages over a human potentially. It can have multiple vision systems and multiple radars and other detection systems. It can have fantastic mathematical calculation ability and it has access to enormous databases of information on the highways and streets we don’t have. In principle a car could be far more aware of the dynamics around the car than a human ever could, but in practice our AI gives us an edge because we use it to predict the motion and behaviors of everything and everyone around us. This is extremely difficult and problematic for computers to do today.
This means we are processing a lot more than simply the relative position of vehicles around us. My point is that everyday driving at high speed involves a lot of assumptions and AI like experience to help us navigate and it is unclear these things can be programmed in a straightforward if then else manner.
All the possibilities and the complexity and all the information requires several things which are quite difficult with today’s technology. How do you integrate conflicting information from multiple sensors into one comprehensive decision path?
Our roads are not perfect, our signs are not perfect, people are not perfect, information is wrong in databases and there is outage of networks and slow response times. Sometimes computer vision fails to find or recognize things humans see easily.
Humans have our eyes primarily to drive. If we had a second source of information we would be distracted by that information and need a more complex rule base to decide when to trust sense A, when to trust sense B and when to use both A and B to figure out the path.
Our computer vision systems are still evolving. They don’t always recognize things and they are especially bad if the things are distorted or noisy. They recognize things that aren’t there. Imagine a car is zooming down the highway and it is blinded momentarily by the sun or the reflection of another vehicle. It could go crazy and go “I don’t know what to do” or it could use GPS information for a short time to estimate the road and path ahead for a short time. However, if the GPS information is flawed it may not make the correct decision. It may have LIDAR and be able to sense no cars are immediately moving closer or obstructions seem apparent.
If you have driven in large metropolitan areas much you know there are a lot of scary situations that happen in NYC that someone in Mountain View, Ca would never see.
I think even the Google technology is going to need some serious tweaking and more testing before it can handle some cities. Take Boston for example. 🙂 In some cities of the world I have driven people routinely flaunt speeding lights and every kind of signage.
In other places I have seen people using a 2 lane road more like a 5 lane road with cars routinely using the breakdown lane as if it were part of the road. I have been driven in SriLanka for instance where animals, motorcycles, regular bikes, tut tuts and every imaginable form of transit coexists in one road coming in and leaving randomly. People cross the street as well as animals as if the road was a passenger thoroughfare.
While these may all be programmable it is likely that by the time we figure that out they will have made changes like the rest of the world to improve the driving situation.
Vision Systems are still evolving
A simple thing is seeing at night. Our computer vision systems are much worse at low light than during daylight. They rapidly lose resolution with lower light. This is not as much the case with human vision because our eyes are something of a miracle of quantum mechanics. A single photon of light is detectable by the human vision system using a quantum trick called quantum tunneling. Cameras are not nearly as sensitive. Night vision systems are better but depend on infrared energy which is heat energy. They will not see anything much better unless it is emitting differential amounts of heat than something else. Generally that is pretty useless on roads. Radar can see physical objects but they tend to respond to bulky objects better and a small obstruction on the road is not as reliable.
A car vision, radar or other sensing system can have false readings and may detect something small and swerve when nothing was there. Our brains process visual information and other information to detect what objects are and so we recognize a piece of cardboard on the road as insignificant whereas a car vision system might think it was a rocket (if a rocket was printed on the surface.)
So, it is likely all these systems will not work in low light or fog or other conditions like heavy rain, snow or anything but perfectly clear sun.
I realize there are cases where the vision systems may be better than a human. If you have ever missed a stop sign and gone through accidentally you will know what I mean. Do you watch every speed limit sign to see the limit? Are you looking at the road every second? I personally would like to have a second set of eyes that was catching things I missed. So, I welcome the help but I know that vision systems today are not nearly at a state that they can be completely dependable especially in more marginal conditions.
We may have to go eventually to the NASA approach to reliability
A car could have other cameras at different angles to switch to. Like NASA it may have 5 cameras. 4 of the cameras operate independently and vote on the conclusion of what they see. If they agree 3 to 1 or 4 to 1 then fine. Do what the majority thinks. However, if the vote is 2 to 2 then NASA determined the best path was to depend on the results of a 5th sensor based on a different technology, different programming and it would assume that something in the other 4 camera system had failed and throw out that whole systems answers. NASA found this was the most reliable way to build systems where human lives depended on the answer.
This may be overkill but as the cost of sensors comes down maybe such a system would be feasible but our sensors suffer from other even more intractable problems.
What if there is construction on the road today. A good visual system will pick up the cues for the construction. However, humans aren’t always diligent in putting signs up or putting them in the right places. They may simply say construction ahead and you have to slow down and figure out how to navigate. A car might be able to do this is programmed with some pretty clever algorithms but maybe not.
There are lots of roads and intersections which are not precisely where the GPS maps indicate they should be. A human mind sees take a right ahead and figures out this driveway is not the right turn intended.
Solution – Safe Zones
One solution to some of these problems would be to have regions of the US that were certified safe for autonomous driving. These areas would be strictly enforced to have good GPS information, have consistent signage in all situations that would be readable by computer. Lane markings may be enhanced to have special reflective properties to radar or vision systems.
A car could know when in an enhanced certified autonomous driving region that things were standardized and markings and information provided was reliable. Eventually more and more areas would meet the requirements. Outside of these regions the autonomous driving would have much stricter limits what it could do. Possible there would be “less safe zone” or “driver must drive in this zone or at this time”
We need to combine what the autonomous vehicle can do and what the community can do to improve the information and reliability. This will mean both changes in infrastructure as well as behavior
We need to change our behavior and be aware that anything we do to a road or people we have on the road, changes in signage or physical changes need to be reported into automated systems so the autonomous cars in the area will be able to plan in advance. Signs have to be standardized and placed more reliably. Every inch of our highways and roads need to be examined for consistency and predictability to make the autonomous vehicle safer and everyone needs to be aware that they can’t assume that if they
This is not terribly hard. It is a matter of signs and process but we must be more aware of the importance of this and more consistent. It’s not going to cost trillions to do this. Maybe billions to get all the roads up to a higher standard and our information systems and reporting systems automated and accurate.
Doing this will drastically improve the safety of everyone whether driving autonomously or not.
Self-Driving car Conclusion
The goal of fully autonomous vehicle operation is hampered by several things that will take a concerted effort to solve in the next decade.
- Improvement of vision systems to faster more reliable and more light sensitive capabilities, more redundancies of sensors.
- Consistency in our roads, signage, GPS information, processes for reporting everything that happens on roads to make the information accurate and more accessible to computers.
- Communications between vehicles electronically to inform each other of desired actions or state would help in complicated situations
- Lots of “experience” gained so we can improve these systems in tandem with the cars intelligence.
Until we can solve these issues there will be the need for humans to still drive in many situations. This could take 10 years or 20 years or longer depending on our appetite and ability to solve these problems and fix our infrastructure.
So, while many of us would love to have the self-driving car tomorrow it is likely to seem far off goal for the next 5 years to 10 years while we work the kinks out, the systems become cheaper and more reliable, that we learn enough to make them be able to work in controlled areas.
Ironically when the vast majority of cars are self-driving the situation gets much easier. When the cars will know what other cars are doing and can work together they theoretically could eliminate all accidents and drive even faster than today. It may be common to have cars driving 200mph on the roads in 20 or 30 years if all the cars were smart speed limits could be raised or even eliminated. Concerns about drunk driving could be removed.
The future is nonetheless still looking bright and not that far off
A lot people’s perception about quality of life revolves around waiting in lines and especially the blight and problems of our transportation system. If we solve energy problems and have bigger batteries by then it is inevitable we will have a lot of the same public infrastructure we have now but it will be more efficient in cost and time to get around.
With that will come other improvements.
It won’t be hard to automate finding and parking automatically. It is inevitable that in 10-20 years many parking spaces will be mapped. Your car will be able to find parking in advance of getting someplace and drop you off and park itself. While the future isn’t full of tubes as depicted in many scifi movies it will be a much less painful infrastructure for getting around than we’ve ever had.
It is inevitable that if we have sufficiently cheap energy source there is no reason with mostly autonomous vehicles and no drunk drivers that highway traffic speeds can move up to 100mph or even higher to 200mph for long distance. The efficiency of driving a car at 200mph is not very good but trucks operating in tandem might be able to achieve significant efficiency even at higher velocity. My belief is that the future will not go down this path. By the time we get fully autonomous vehicles another technology will replace our longer distance needs.
The future looks bright indeed and I believe that future is 20 years off to get to this ultimate spot where most cars are self-driving and parking is automated and the number of fatalities in cars has dropped dramatically possibly by 90%.
The hyperloop is based on the idea that if you can evacuate most of the air from a tube the normal issues of speed and sonic booms go away. There would be the ability to accelerate much faster than with wind resistance and the transportation efficiency would jump dramatically as air resistance is a very large part of the overall inefficiency of typical transportation systems.
An even more interesting thing turns out to be possible. When you arrange magnets around a tube just right the metal car travelling inside the tube experiences a significant force that lifts the vehicle. This is possible without any external power application other than the power to move the vehicle.
The existence of this effect means that once set into motion a vehicle in one of these pods could literally float on the air left in the pod and have almost zero resistance with the surfaces as well as with the air.
This would mean the cost of going 5 miles or 5000 miles would be incredibly low compared to traditional transportation to the point of essentially being free. It would easily be possible for the cost to go from San Francisco to London would be less than the cost of a subway ticket in any city.
The physical problems with achieving hyperloop are infinitely less than trying to get autonomous vehicles working ubiquitously
Once such hyperloop tubes were set up and the magnets in place the tubes would be a tremendous resource because they would require only making sure they were relatively sealed (a perfect vacuum is not required) and the tubes should work forever at zero additional cost. The pods running in the tubes would be upgraded periodically but their cost of maintenance would be almost zero as well.
It is easy to imagine transport at thousands of miles per hour for virtually free throughout most of the world. The economic efficiency of this form of transport is so compelling that unless we are talking small distances it will make sense to use hyperloop for all future long distance transport. This would also mean shipping costs would drop dramatically.
It’s up to us how fast this can happen
This will take longer than 20 years. The first hyperloop may take 20 years to build at the current rate. It will probably take 60 years to 80 years to deploy hyperloop all over the world but once going these things will multiply like hotcakes. The economics of them is unassailable.
Even though we could have cars going 200 mph on freeways it will be grossly inefficient to do that and pointless unless you wanted to do that. It is likely that full VR projection from the outside to the inside of a hyperloop car will make being inside the hyperloop as visually exciting as being in a car.
Competing with this future is the need to travel at all. VR technology may make transporting oneself everywhere far less needed. However, we will still want to transport lots of “stuff” and so the benefits for cargo transportation would be there.
How all these factors play out is hard to figure out but the sheer economics of these things and the technical advances in the pipeline foretell that this is likely our future.
I would like to travel for free and get to the future I’ve described above as fast as possible. I believe autonomous vehicles are coming along very fast. Once people see the economics of building hyperloop it will explode in a thousand places.
What are the social consequences of these phenomenon? Autonomous driving? How will it change school, people drinking and partying more? How will it change the commute to work or travel? Will more people choose to do a car trip when they can lay back in the comfort of a car designed for comfort and play rather than safety and density?
How will free transportation over long distances change where people live, how they work, where they travel?