Building electric cars and reusable rockets is fairly easy. Building a nuclear fusion reactor, flying cars, self-driving cars, or a Hyperloop system is very hard. What makes the difference?
The answer, in a word, is experience. The difference between the possible and the practical can only be discovered by trying things out. Therefore, even though the physics suggests that a thing will work, if it has not even been demonstrated in the lab you can consider that thing to be a long way off. If it has been demonstrated in prototypes only, then it is still distant. If versions have been deployed at scale, and most of the necessary refinements are of an evolutionary character, then perhaps it may become available fairly soon. Even then, if no one wants to use the thing, it will languish in the warehouse, no matter how much enthusiasm there is among the technologists who developed it.
It’s well worth considering what makes a potential technology easy or hard to develop, because a mistake can lead to unwise decisions. Take, for instance, the International Thermonuclear Experimental Reactor that’s now under construction in France at an estimated cost of US $22 billion. If governments around the world believe that this herculean effort will automatically lead to success and therefore to near-term commercial fusion reactors, and if they plan their national energy strategies around that assumption, their citizens may very well be disappointed.
Here I present a short list of technology projects that are now under way or at least under serious discussion. In each case I’ll point out features that tend to make a technology easy or hard to bring to market.
Not Much Needs to Change
Electric cars are a relatively easy technology because cars have been mass-produced for more than a century. We have more than 100 years of experience engineering and manufacturing windshield wipers, brakes, wheels, tires, steering systems, windows that can go up and down, car seats, chassis, and much more. We have more than 20 years of experience making digitized drivetrains.
On top of that, we already have a whole infrastructure for driving, including roads, parking spaces, safety standards, auto insurance, and government licensing of both the vehicles and the drivers. So to go from internal-combustion-engine cars to electric cars, you don’t have to invent everything from scratch and then figure out how to deploy it at scale.
True, to mass-produce electric cars at a competitive price, with good range and reliability, you have to be very clever—you need good batteries, for one thing—and well capitalized. But there is much that you do not need to change. For that part, there are plenty of people who have worked on the relevant components for decades and plenty of expertise for building and assembling the components. Electric cars constitute a new technology, but not an unreasonably hard one.
Likewise, reusable rockets may sound revolutionary, but here again there is plenty of prior art. All liquid-fueled rockets derive from the V-2 rockets that Wernher von Braun built for Hitler. The V-2 had high-flow turbopumps (433 kilowatts!) that circulated the fuel to cool parts of the engine, and it carried its own liquid oxygen so that it could fly above the atmosphere. The first flight of the V-2 happened just over 76 years ago. And it went on to be mass-produced, albeit with slave labor.
Since then, over 20 different families of liquid-fueled rockets have been developed around the world, some of those families coming in hundreds of different configurations. Soyuz rockets, a 52-year-old family, all lift off with 20 liquid-fueled thrust chambers burning. In the Delta family, the “Heavy” variant of the Delta IV has three essentially identical cores side by side, each being a first stage of the earlier, single-core Delta IV.
The technology for soft landing on Earth using jet engine thrusters has been around since the 1950s, when Rolls-Royce demonstrated its “flying bedstead.” The following decade came the Harrier fighter jet, which also could take off and land vertically. In 1969, a manned rocket—the lunar module—vertically landed people on the moon. And in the 1990s, McDonnell Douglas built the single-stage Delta Clipper Experimental, or DC-X, rocket, which took off and landed vertically half a dozen times at the White Sands Missile Range, in New Mexico.
Today’s reusable Falcon rocket, from SpaceX, uses grid fins to steer the first stage while it is returning for a soft landing either at the launch site or on a recovery barge. The theory behind grid fins was developed in Russia in the 1950s by Sergey Belotserkovskiy, and rockets equipped with these fins have been used since the 1970s for missiles and guided bombs and for the emergency escape system in crewed Soyuz capsules.
I am by no means saying that developing electric cars or reusable rockets is not brave, hard, and impressively inventive work. This work does, however, build on large bodies of prior work and on existing physical and business infrastructure, all of which increase its chances of success. There are known solutions to many, though not all, of the problems that will arise. And so, with some confidence, we can make estimates about these technologies being technically successful and deployable at scale.
Haven’t Been There, Haven’t Done That
Completely new ideas, however, are much harder to evaluate. When or even whether they will succeed isn’t clear, no matter how logical these ideas may seem at first glance.
The thermonuclear fusion reactor is an example of something that’s old and yet still barely closer to realization than it was when completely new. It has been under development since the 1950s, a time at which we already knew that sustained nuclear fusion “works.” That’s how the sun shines, after all. Humans first produced a short-lived nuclear fusion reaction with the detonation of the “Ivy Mike” hydrogen bomb 66 years ago. Back in the day, futurists confidently predicted that nuclear fusion could be harnessed to produce electricity within a reasonable period of time, but it hasn’t happened yet. I doubt that many people today would believe any particular predicted date for practical fusion power generation.
For sustained nuclear fusion, extraordinarily hot gases must be contained at extraordinarily high pressures. No physical container can withstand such temperatures and pressures. Instead, vastly strong magnetic fields must be used as a nonphysical container. The magnetic fields necessary have turned out to be very difficult to generate and control, and no one is confident that we are close to solving all the engineering problems, even after 50 years of work on them.
I need not argue further: It is a really hard problem.
The flying car is another old dream that’s back in vogue. Originally, the dream was that you’d drive down the road just as far as needed to find clear airspace, fly nearly to your destination, then land and get back on the road for the last leg of the journey. Your flying car would let you leap over road congestion and cruise at a faster speed while doing so. The dream was never realized, but now a baker’s dozen of startups are pursuing the idea, and the number of engineers working actively on it has increased enormously in just the past dozen years.
The problem is hard because a flying car combines two completely different engineering regimes. It’s not straightforward to engineer something that can both fly thousands of meters above the ground and also fit within the narrow space constraints that a road and highway network imposes on conventional automobiles, all the while meeting the diverse safety and efficiency requirements of flight and ground transport. Optimizing for one regime means scanting the other.
It’s not surprising then that what today’s startups call a flying car is generally something rather different: They are working on point-to-point flying vehicles, mostly electrically powered ones that some are claiming could be piloted by ordinary people, without significant training. What the vehicles typically don’t have are road wheels, which means you need some other means of getting to wherever your flying car is parked, and then once you land, you need a way to get to your final destination.
While this variant on the flying car doesn’t have to drive on roads, it has other problems: The vehicles must somehow get recharged or refueled. As ultralight airplanes, they aren’t allowed to fly over structures, a restriction that will impede their usefulness for commuting. Amateur pilots, with next to no training, will still have to abide by air-traffic-control rules and pass muster with the insurance companies.
Not a single public demonstration flight has yet occurred, or even been claimed. The rules, regulations, and insurance have not begun to be worked through. Please do not hold your breath for any flavor of this flying-car dream to come true.
Obstacles are closer than they appear
The self-driving car is arguably the single most anticipated technology right now. Here the difficulty lies in attempting something that has no real precedent.
Last year I wrote in this magazine on one aspect of the problem: the unexpected consequences that self-driving cars might have on human behavior [see “The Self-Driving Car’s People Problem,” August 2017]. I pointed out that pedestrians and the drivers of other cars might find autonomous cars a tempting target for antisocial behavior. I also noted that the owners of self-driving cars may use them in ways that they would never use a regular car, perhaps succumbing to antisocial behavior themselves.
Another problem is what are called edge cases, which involve robotic cars bumping up against the limits of their capabilities. Some of those limits are not known ahead of time. Examples that we do know about include conditions under which the car must read and interpret temporary signs, such as those warning of road construction; conditions where it’s appropriate to skirt the letter of the law; conditions where a ride-hailing service must figure out how much control the passenger is allowed to have; and conditions where the car must determine what to do when human drivers can no longer communicate with other human drivers, for instance while negotiating entry into a lane.
Driverless cars will not simply replace cars that have human drivers. Instead, we’ll install special lanes, even geofencing the self-driving cars into lanes or entire roads, of their own, to protect them from human-driven cars and vice versa. Also, we’ll change the norms for where it’s acceptable to pick up and drop off people, where to park, and many other things.
It may seem that self-driving cars have suddenly made enormous advances. However, if you look back, the progress has been incremental over the 32 years since Ernst Dickmanns and his colleagues at the Bundeswehr University Munich had their autonomous van drive on a public highway.
It is only in the past year that actual autonomous cars without a safety driver have begun to ply public roads—Waymo’s ride-sharing project, near Phoenix. And Waymo (a subsidiary of Google’s parent company, Alphabet) is just at the demonstration stage with this effort.
The price of the sensors still needs to come way down, and how the cars will be used still needs to be worked out. We’ll need changes in safety regulations and in how we assign legal liability. And for the laws to change, attitudes must change.
The true proof of self-driving cars will come when they graduate from the status of a science experiment to that of a commercial enterprise—that is, when the makers of these cars actually begin to turn a profit on sales, either to individuals or to fleet operators. At first the cars will operate in restricted geographies and markets, such as malls, industrial campuses, and other places where human-driven cars are not allowed. Perhaps they will be restricted to certain times of day and certain weather conditions. The various problems of the self-driving car will be solved—eventually. But it will all unfold more slowly than the enthusiasts think.
No Component Is Too Hard, but All Together They’re a Bear
The Hyperloop is yet another technology that’s harder than it seems. The idea is to build a partially evacuated tube through which capsules full of people or cargo can be accelerated, either with externally applied air pressure or through the use of magnetic induction coils. The concept has attracted the imagination of many entrepreneurs and capital from many backers, and yet nothing like it has ever been demonstrated, let alone operated at scale.
Figuring out how to develop the tube itself—an ultrastable, airtight cylinder that goes on for hundreds of kilometers in a very straight line—is one problem. You also need to engineer the capsules, which will travel at nearly the speed of sound while carrying people. The sealed capsules will require an entirely self-contained life support system. Any stations where a capsule won’t stop will also needed to be sealed as the capsule passes by, while any station where passengers are getting on or off will need to allow them to do so. Emergency procedures must be worked out, for instance, to extract passengers from a capsule that gets stuck a hundred kilometers from the nearest station and then to remove the capsule so that the tube can be reopened. You’ll need some way to communicate with the capsules, one that can penetrate what might otherwise be a pretty good Faraday cage.
Seats and restraints must be developed for the safety of the passengers and for their sanity as well—it may not come naturally to sit wedged in a seat in a windowless chamber under hard acceleration. The entire system will need to be protected against earthquakes, as well as the subtler displacement of the tube as the tectonic plates underneath it shift by a centimeter or two. And don’t forget getting rights to the land for the routes, insurance (including figuring out how the Hyperloop’s own insurance interacts with the policies of individual passengers), business models, and on and on.
You could argue that no one aspect of the Hyperloop is too hard to master, but together they constitute a hard problem indeed. Many new technologies and designs must be developed from scratch and then demonstrated. At this point, they haven’t even all been enumerated.
Then there’s the completely different problem that comes after all the technological challenges have been met and definitively and successfully demonstrated. And that is the psychological problem: It’ll be a hard sell to coax passengers into those windowless, high-speed systems, at least at first. Finally, even a safe and functional Hyperloop may not pay back the money invested in it for a lot longer than its proponents expect.
Sometimes, the Possible Just Takes a Little Longer
Sometimes progress on even an easy technology can be slowed to a crawl despite there being no obvious impediment. One of the best examples is in how we organize addresses on the Internet.
Internet Protocol version 6 uses 128-bit addressing, up from the 32 bits of the previous protocol, IPv4. It thus raises the number of potential unique addresses for all the devices on the network from just 4 billion to a number that’s greater by the ridiculously high factor of 7.9 × 1028. Engineers developed the new version in the 1990s after it became clear that far more devices would be joining the network than had been anticipated—not just computers but electricity meters, industrial sensors, traffic sensors, television sets, light switches, and so on. Plenty of ingenuity was expended in cramming far more than trillions of devices into that puny 4-billion-device address space. And yet although IPv6 was fully defined by 1996, it still hasn’t taken over.
In 2010, the target date for the switchover from IPv4 to IPv6 was 2012. In 2014, fully 99 percent of all network traffic was still using IPv4. By the end of 2017, network traffic running on IPv6 ranged from under 2 percent (going through the Amsterdam Internet Exchange) to just over 20 percent (for users of Google services). Clearly IPv6 is a work in progress—slow progress.
In pointing out the differences that make one technology harder than another, I am not preaching technological defeatism. I’m only suggesting that we properly gauge the difficulty of whatever we are told could be the next big thing. If the idea builds on practical experience, then guarded optimism is in order. If not, then not. Hope is a scarce thing; we shouldn’t squander it.
This article appears in the November 2018 print issue as “How to Bet on Tech.”