## 2021-03-27

### Nuclear power is good

(Alternative title: burning things considered harmful)

If you want usable energy, you need to use the forces between particles.

The weakest force is gravity, but if you happen to be near a gigantic amount of material (e.g. the Earth) with an uneven surface that has stuff flowing down it (e.g. water in a river), we can still use it to generate power. This insight gives us hydropower, which delivers about 16% of the world's electricity. The main downside is that because of how weak gravity is, dams have to be large and environmentally disruptive to generate useful power.

Moving to stronger forces, we have chemical interactions between atoms. In the form of burning fossil fuels, rearranging chemical bonds produces 66% of the world's electricity. The main downside is how weak chemical bonds are, and therefore how much matter has to be processed (i.e. burned) to produce energy. A lot of matter means a lot of waste products. Despite decades of work on possible safe waste-management strategies (e.g. carbon capture and storage), we still outrageously keep dumping over thirty billion tons of carbon dioxide into the atmosphere every year, with massive effects on the climate that will potentially last thousands of years, while also producing a long list of other harmful waste products that kill a lot of people per year.

Thankfully, atoms aren't atomic: we can rearrange atoms and get energy densities that blow puny chemistry out of the water. Currently 11% of the world's electricity comes from directly doing this. We're still playing catch up to God, who, in His infinite wisdom, saw it fit to create a universe where just about 100% of energy production is nuclear.

Our nearest God-sanctioned nuclear reactor is the sun. Harnessing the sun's light and heat gives us another 1% of the world's electricity; a slightly more indirect route where we first wait for the sun's heat to stir up the air gives us another 3.5%. An even more indirect route is letting the sun's light fall on plants so that they create chemical bonds that we can burn for power; this gives us another 2%. The most indirect route of all is to use the chemical bonds created by sunlight that fell on extinct plants hundreds of millions of years ago, which is what we're really doing when we burn fossil fuels. So actually it's all nuclear, with the only difference being how many hoops you jump through first.

The current state of nuclear power is that we can harness only fission (splitting atoms) for controlled energy production. Fusion (combining atoms) is potentially an even better technology: it requires less exotic materials, produces less dangerous waste, and is literally star-power. However, it takes extreme energies to get power out of fusion, and the only way we've found how to do that is to blow up a (fission-based) nuclear bomb in a very controlled way that squeezes the stuff we want to fuse to create an even bigger bang. Technically we could use this for power – say, we build a massive underground chamber where we set off hydrogen bombs (the common name for a bomb that uses nuclear fusion) every once in a while to vaporise vast amounts of water into steam and then drive a generator – but let's just say there would be some difficulties. (Though, surprisingly, mostly economic and political ones rather than technical ones – this idea was seriously studied in the 1970s as Project Pacer.)

Controlled fusion power is in the works, but it's the poster child for technologies that are always twenty years away. At the moment scientists are playing around with lasers that have 25 times the power of the entire world's electricity generation (though only for a few picoseconds at a time) and magnets almost strong enough to levitate a frog* to bring it about, but don't expect commercial fusion power in the next decade at least.

(*Levitating a frog takes a field of about 16 Teslas, according to research that won an Ig Nobel Prize in 2000, compared to ITER's 13 Tesla field.)

Fusion is definitely a technology that we should develop. However, as J. Storrs Hall writes in Where is my flying car? (my review here):

"As a science fiction and technology fan, for most of my life I had been squarely in the “just you wait until we get fusion” camp. Then I was forced to compare the expected advantages fusion would bring to the ones we already had with fission. Fuel costs are already negligible. The process is already clean, with no emissions. Even though the national [US] waste repository at Yucca Mountain has been blocked by activists since it was designated in 1987 and never opened, fission produces so little waste that all our power plants have operated the entire period by basically sweeping it into the back closet."

We have already invented a miracle clean power source. And, surprise surprise, we should really use it.

## The human case for nuclear power

Every year, there are almost five million deaths attributable to air pollution, a bit less than 1 in 10 of all deaths in the world, or one every six seconds. Since it's a bit tricky to know what counts as an "attributable death" in the case of some risk factor, here's another measure: almost 150 million years of health-weighted life are lost every year because of air pollution. The health effects of air pollution are right up there with the other biggest killers like high blood pressure, smoking, and obesity.

The biggest causes of air pollution are energy generation, traffic, and (especially in poor countries) heating. Getting global averages for power generation deadliness is hard, but doing some very rough estimation, more than one-tenth but less than one-third of air pollution deaths are directly related to power generation, for a total number in the hundreds of thousands per year. Imagine three Chernobyl-scale disasters a week, and you're in the right ballpark.

(There is major disagreement over the actual Chernobyl death toll. When making comparisons in this post, I use the number 4000. About 30 people died directly during the disaster; several thousand may die in the long run according to the best consensus estimates, though if you assume the contested linear no-threshold model (which seems to be the main crux of the debate) you can get numbers in the tens of thousands. If you want to be maximally pessimistic, you can multiply Chernobyl impact comparisons by 10, but you'll find this doesn't materially change the conclusions.)

Which power sources cause these deaths? There's some disagreement over the exact numbers, but here's a chart for European energy production from Our World in Data:

(One terawatt-hour (3.6 petajoules) is roughly the annual energy consumption of 20 000 Europeans.)

The chart above has European numbers. In particular for fossil fuel sources, there's a lot of country-specific variation due to environmental regulations and population density: the paper that the above chart is largely based on mentions 77 deaths/TWh as a reasonable figure for a regulation-compliant Chinese coal plant, while this article says that 280 deaths/TWh is possible for coal.

Why do solar and wind produce any deaths at all? Both occasionally involve dangerous construction work (rooftop solar / tall wind turbines). In fact, if you look at recent decades (i.e., not including Chernobyl) and use the low-end estimates, solar and wind are deadlier than nuclear.

The estimates for hydropower can also swing a bit depending on whether or not you include the deadliest electricity generation disaster in history: the 1975 Banqiao Dam failure, which may have killed hundreds of thousands of people. Since 1965, hydropower has produced about 130 000 TWh; depending on which death toll estimate you believe, Banqiao single-handedly raises the deaths per TWh for hydropower by between 0.2 and 2. Compare this with nuclear power, which has produced about 92 000 TWh over the same timeframe; the long-term death estimates for Chernobyl add 0.04 to the deaths/TWh count for nuclear.

(The total generation numbers are based on the raw data behind this and this graph, which you can download from the links. The nuclear number in the above chart is based on this paper, which Our World in Data says already includes Chernobyl, though I can't see where they add that in.)

The bottom line is that hydropower accidents are more common, more deadly, and higher variance than nuclear accidents, even though both power sources have produced comparable amounts of energy in recent decades.

Okay, actually that isn't the real bottom line. The real bottom line is this: when it comes to the human impacts of electricity generation, there are things that involve burning (fossil fuels & biomass), and then there is everything else, and the latter category is much much better. Also, if you absolutely must burn something, do not burn coal.

What has nuclear specifically done so far? One study finds that it has saved 1.8 million lives by reducing air pollution, or about 4 years of the world's current malaria death rate.

What could it have done? Until the mid-1970s, the adoption of nuclear power was accelerating. Assume this trend had continued until today, and nuclear had replaced fossil fuels only (an optimistic assumption, but one that doesn't change the numbers much because renewables are a pretty small percentage). Under these assumptions, one study estimates that nuclear would now account for over half of the world's energy production, and a total of 9.5 million deaths would have been avoided – as much as if you saved everyone who would otherwise have died of cancer in the past year. Even if nuclear adoption had only been linear, 4.2 million deaths could have been avoided, the same number as saving everyone who has died in war since 1970 (the war deaths number is from the raw data behind this chart).

Therefore: in terms of the number of lives saved, keeping the nuclear power industry growing would have very likely been at least as good as achieving world peace in 1970.

Since these numbers are enormous, and involve difficult-to-estimate unknowns, here's something more concrete: Germany's decision in 2011 to get rid of nuclear is costing an average of 1100 lives per year (working paper; article).

## The environmental case for nuclear power

Climate change is a big problem, but the scale of it as an environmental problem is better known than the scale of air pollution as a health problem, so I won't go into the statistics on its impact.

Nuclear power is obviously good for the climate. Here's a chart, based on this, which is summarised in a more readable format here:

The black bars span the range between the minimum and maximum numbers. The red dot is the median.

I've converted the numbers from the traditional grams of CO2 equivalent per kWh to tons of CO2 equivalent per TWh, to be consistent with the death rates graph above, and for easier conversion to national/international CO2 statistics (which are generally expressed in tons of CO2 – unless its tons of carbon, in which case you divide by the ratio of carbon's mass in CO2, which is 12/44 or about 0.27).

(If you're wondering where hydropower is: it's median is right around concentrated solar, but in some cases, especially in tropical climates, the reservoirs created by dams can release a lot of methane, making the maximum CO2-equivalent emissions for hydropower over twice as bad as coal and, more importantly, completely ruining my pretty chart.)

So far, the use of nuclear power is estimated to have reduced cumulative CO2 emissions to date by 64 billion tons, a bit less than two years of the world's total CO2 emissions at current rates. The same study linked in the previous section estimates that, had nuclear power grown at a steady linear rate, this number would be doubled, and if the accelerating trend in nuclear power adoption had continued, there would be 174 billion tons less CO2 in the atmosphere. We would have saved more emissions than we would have if we had made every car in the world emission free since 1990.

## The problems

In Enlightenment Now (my review here), Steven Pinker writes:

"It’s often said that with climate change, those who know the most are the most frightened, but with nuclear power, those who know the most are the least frightened."

So why aren't the arguments against nuclear power enough to frighten those who know about it?

The short version: more nuclear power would save millions of lives from air pollution and be a big help in solving climate change. When these are the benefit, you need a hell of a drawback before the scales start tilting the other way.

The long version:

(Radiation units are confusing. Activity, straightforwardly defined as the number of atoms that undergo decay per second, is measured in becquerels (Bq). The amount of radiation energy absorbed per kilogram of matter is measured in grays (Gy), which therefore have units of joules per kilogram. Measuring biological effects is harder, because the type of radiation and what tissue it hits both matter. If you adjust for the type of radiation by multiplying the absorbed dose in grays by some factor (scaled so that gamma rays have a factor 1), you get something called equivalent dose, which is measured in sieverts (Sv). If you also adjust for which tissue type was hit by multiplying by more estimated factors, you get effective dose, which is also measured in sieverts. If you want to get a sense of scale for radiation dose numbers, here's a good chart and here's a good table.)

In normal operation, a nuclear power plant produces significantly less radiation than a coal power plant (this is because everything radioactive is contained in a nuclear power plant, while coal power plants pump fly ash into the air). Neither is a significant dose.

In accidents, nuclear power plants can release insane amounts of radioactivity. Insane amounts of radiation are dangerous. However, the reaction to radiation risks is often out of proportion to the true risk – the Fukushima evacuations are considered excessive in hindsight, as argued in this study, though you probably don't need to make a study to guess it from this chart:

(In the long run, some more cancer deaths are expected to trickle in.)

It is critically important to remember the above statistics on health effects, and not let yourself be biased by vivid stories about horrible individual events. The fear of nuclear accidents is similar to the fear of flying rather than driving: statistically one is much safer, but one is much easier to fear because when things go wrong, it comes in more story-worthy packages.

In particular: it is not the case that nuclear power is safer only because accidents are rare and therefore get left out of statistics; nuclear power would be overwhelmingly safer than fossil fuels even if there were a Chernobyl going off every year. As I said above, hydropower accidents are more common, more deadly, and higher variance, so any argument based on disaster risk that bans nuclear would also ban hydropower.

### Nuclear proliferation

Nuclear power is good, but nuclear weapons are bad. It would be bad if the spread of civilian nuclear power technology lead to nuclear proliferation. There is some overlap in technology, but neither civilian materials nor technologies automatically lead to weapons. The uranium used in power plants is typically only enriched to 3-5%, compared to more than 85% for weapons-grade uranium and 0.7% in natural uranium (though if you have uranium enrichment infrastructure, you can run it for more cycles than usual and let the enrichment levels slowly creep up – Iran has done this). There are also international agreements that prevent enrichment, and alternative nuclear technologies, like using thorium instead of uranium, with less weapon potential. Finally, a country trying to build nuclear weapons probably won't be stopped by a lack of a civilian industry; consider North Korea.

### Terrorism and war risks

Another risk to consider is that nuclear power plants might be targeted by terrorists, or even by hostile nations, potentially leading to Chernobyl-scale disasters. This is a risk, but it's an acceptable one. Consider what it would mean if "hundreds or thousands of people could be killed if a determined and resourceful hostile actor targeted this piece of infrastructure" were a reason to not build some piece of infrastructure – we'd have to ban skyscrapers, airplanes, dams, water treatment plants, and so forth. Also considering the security that's (rightfully) present at nuclear power plants, it would probably take a 9/11-level of execution to do it, and the observed rate for 9/11-level events over a time interval of length T is, well, 1/T if the interval includes 9/11 and otherwise 0.

It is true that a complex civilisation has a lot of fragile points and someone should be thinking hard about minimising this kind of risk, and that nuclear power plants are a good example because the effects are expensive and long-lasting if an attack is successful. But as an argument against nuclear power, it proves too much.

### Nuclear waste

Nuclear waste is awkward to deal with, but it's far from the worst sort of industrial waste we deal with – consider the over thirty billion tons of carbon dioxide we've dumped into the atmosphere over the past year, or the various horrible things that coal plants spew out that cause dozens of Chernobyl-equivalents per year.

Nuclear waste is not some miracle substance that effortlessly seeps everywhere and kills whatever it touches. Until 1993, countries (mostly the USSR and UK), were dumping nuclear waste into the ocean. This is rightly banned these days, but you can observe that we still have oceans; in fact, the the environmental impacts have so far been negligible except for somewhat higher concentrations of some nasty isotopes exactly at the site.

In general, nuclear waste is a serious problem that has to be solved somehow, but solutions exist (currently, Finland's Onkalo repository is the closest to being operational). Though the timescale is long, it is not different in principle from some existing disposal methods for nasty things like mercury and arsenic.

Is it responsible to leave behind dangerous waste for future generations? It's far more responsible than leaving them with the almost astronomical amounts of CO2 emissions that a single kilogram of uranium prevents.

Future people looking back at our century won't despair about a few warm rocks deep underground. They'll despair at all the silent air pollution deaths, at how far we let climate change get, and at how much sooner we could've reached their living standards had we made better use of our technology. Then they'll travel on nuclear-powered airplanes to distant hiking grounds, and tell scare stories around an (artificial!) campfire about the barbarian past when we burned things for energy and piped the waste products straight into the atmosphere.

### Uranium is limited

First, we have 200 years worth of economically accessible uranium reserves. This is more than for fossil fuels, with the additional benefit that burning through the remaining uranium won't wreck the climate and kill millions.

Second, we have alternatives to uranium, like thorium.

Thirdly, there are hundreds of times more uranium dissolved in the oceans than there is on land (and this uranium exists in equilibrium, so if you take it out, more will leach out of the seabed to replace it, a fact that might lead a pedant to call nuclear power renewable). Even though the concentrations are tiny, because of the energy density of uranium, at modern reactor efficiencies there's still half a megajoule of usable nuclear energy in the uranium in a single cubic metre of seawater, enough to power the lightbulb in my room for over five hours. As a result, extracting it is a project that is taken surprisingly seriously, and is surprisingly close to being economically viable, though some people are very skeptical.

### Nuclear power is unnatural

Wrong: a few billion years ago a spontaneous natural nuclear reactor ran for a few hundred thousand years under what is now Gabon.

Using the best estimates for its running time and power output, even if this is the only natural reactor that ever formed, the energy it produced is several times higher than that of all human civilian nuclear power to date (both numbers are in the hundreds of petajoules range). Of sustained nuclear fission energy in our planet's history, more has been natural than artificial.

## Nuclear is overpowered, so where is it?

Nuclear power is an almost overpowered technology. The reason why comes down to physics: an energy source based on nuclear reactions has extreme power density, and, all else being equal, the higher your power density, the less fuel you need, the less waste products you produce, and the cleaner your power plant is overall. Not surprisingly, nuclear power turns out to be – along with solar and wind – the cleanest and safest power source we have.

In Where is My Flying Car?, J. Storrs Hall gives some vivid facts to demonstrate the power and efficiency of nuclear: a wind turbine uses more lubricating oil per energy generated than a nuclear power plant uses uranium, and while the 7.5 TJ of energy a Boeing 747 burns through during a flight weighs 200 tons and costs a third of a million dollars when delivered as chemical fuel, getting the equivalent energy from nuclear takes 100 grams of reactor-grade uranium and costs 10 dollars.

So where is it? The simple reason is that it's either illegal (like in Italy), being phased out (like in Germany), or highly regulated and/or expensive. It wasn't always so:

 Source: Where is my Flying Car?, by J. Storrs Hall.

The above graph shows the price per kilowatt of US nuclear power plants. The green line is the trend line before the Department of Energy was established in 1977. Note also that the Three Mile Island accident was in 1979, and, despite no one being hurt, this was a turning point for the US nuclear industry.

When the price of a technology starts increasing, it's not the natural learning curve of the technology at work. It's a regulatory choice. And while you obviously should regulate nuclear power, we're not doing it right.

J. Storrs Hall explains the cost increases:

"Nuclear power is probably the clearest case where regulation clobbered the learning curve. Innovation is strongly suppressed when you’re betting a few billion dollars on your ability to get a license to operate the plant. Besides the obvious cost increases due to direct imposition of rules, there was a major side effect of forcing the size of plants up (fewer licenses); fewer plants were built and fewer ideas tried. That also meant a greater cost for transmission (about half the total, according to my itemized bill), since plants are further from the average customer."

There is some hope that the tide is turning. New startups like NuScale are working on small modular reactors that might greatly reduce prices. Of course, in addition to difficulties with funding, and the not-so-easy task of building a literal nuclear reactor, they've spent years jumping through regulatory hurdles and are not expected to produce power until 2029. So-called fourth-generation reactors are also being worked on, and there's always the hope we eventually get fusion.

But we're not going to get the benefits of cheap and plentiful nuclear power unless we stop treating it like it's the Antichrist.

Hall, never one to pass up the opportunity for a dramatic touch, quotes John Steinbeck's The Grapes of Wrath to sum up the sadness of our attitude to nuclear power:

“And men with hoses squirt kerosene on the oranges, and they are angry at the crime, angry at the people who have come to take the fruit. A million people hungry, needing the fruit—and kerosene sprayed over the golden mountains.

[...]

There is a crime here that goes beyond denunciation. There is a sorrow here that weeping cannot symbolize. There is a failure here that topples all our success. The fertile earth, the straight tree rows, the sturdy trunks, and the ripe fruit. And children dying of pellagra must die because a profit cannot be taken from an orange. And coroners must fill in the certificate—died of malnutrition—because the food must rot, must be forced to rot.”

More generally, human civilisation need to get better at making decisions about technology. We shouldn't deny ourselves safe clean energy, but we should start working on mitigating the harms from actually scary technologies, like nuclear weapons, and make sure that new technologies like biotech and AI are used safely. Oh, and have I mentioned that burning things is bad for climate and health, and we should stop doing it?

## A metaphor

I mentioned earlier that nuclear power and fossil fuels are like flying and driving. One of them is obviously safer, but the other seems scarier because the lizard-derived part of our brains can't multiply. Objecting to nuclear power on safety grounds but tolerating fossil fuels is like texting about how scared you are to board a plane while driving yourself to the airport. Let's make this metaphor more concrete, and hopefully create a memorable image.

The world consumes about 20 000 TWh per year as electricity (about one-eight of total energy use – lots is used directly for transportation and heat). Let's compare this to making a drive across Europe that starts in Lisbon and ends in Tallinn. Each kilometre we travel represents a bit less than 5 TWh of energy towards our 20 000 TWh goal. Let's say walking is wind/solar/geothermal, biking is hydropower, flying is nuclear, and driving is fossil fuels.

(The numbers for fossil fuel related deaths below are significant underestimates of the global average, because, like the chart above, they're based on the European data in this study. Regulations are looser and population densities higher in many developing countries that make up most of the world's air pollution deaths. I was not able to find a good estimate of the global average, and besides, these numbers are terrifying enough as they are.)

First we walk some 450 km, ending north-west of Madrid, and then bike 650 km, just barely taking us into France. We're a bit careless and somehow we've manage to shove a hundred people off wind turbines along the way. Oops.

By this point we're getting tired of walking and biking, but thankfully there's a flight to Paris. The pilot has a bad day and lands on top of a crowd, flattening another hundred people.

We really hate flying, so we refuse all the other offers that the airline companies try to sell us. Instead we step out of the Paris airport, rent a car, and start carelessly careening down the remaining 2600 km.

Gas takes us approximately to Berlin, a distance of about 1000 km. During this entire distance we run over a pedestrian at every block (roughly 1 per 80 metres), killing some 10 000 people in total.

We're in a real hurry to get to Poland, where the traffic rules get even more lenient and we can start burning coal. The final leg of the journey from Berlin to the Polish border is powered by oil and isn't long, but still results in as many lethal hit-and-runs as the entire journey before it.

At the Polish border, we reach coal. From this point on, we text about the dangers of nuclear waste as we mow down one pedestrian every 8 metres for the entire rest of the coal-powered trip to Estonia (also burning some other nasty things too). Driving at a reckless 120 km/h whatever road we're on, we go run through four pedestrians a second – you'll hear a rapid thwack-thwack-thwack-thwack noise as the bodies hit the windshield – but it still takes 13 hours to make the trip. By the time we reach the Lithuanian border, the bodies of our victims, packed as tightly as possible, fill four Olympic swimming pools. Each of the three Baltic countries we drive through before reaching Tallinn fills another one.

Oh, and also every kilometre driven in our car had fifty times the environmental impact of flying.

Thank god we didn't fly: imagine how horrible it would be if another pilot had had a bad day.

The world makes this trip every year to meet our growing energy needs. We're getting fitter and walking a bit longer every year, as we should. But whenever someone suggests flying instead of driving, our collective response is: "What?! But that's so risky!"

Let's fly.

RELATED:

## 2021-03-25

### Technological progress

In this post, I've collected some thoughts on:

• why technological progress probably matters more than you'd immediately expect;
• what models we might try to fit to technological progress;
• whether technological progress is stagnating; and
• what we should hope future technological progress to look like.

## Technological progress matters

The most obvious reason why technological progress matters is that it is the cause for the increase in human welfare after the industrial revolution, which, in moral terms at least, is the most important thing that's ever happened. "Everything was awful for a long time, and then the industrial revolution happened" isn't a bad summary of history. It's tempting to think that technology was just one factor working with many others, like changing politics and moral values, but there are strong cases to be made that a changed technological environment, and the economic growth it enabled, were the reasons for political and moral changes in the industrial era. Given this history, we should expect that more technological progress will be important for increasing human welfare in the future too (though not enough on its own – see below). This applies both to people in developed countries – we are not at utopia yet, after all – as well as those in developing countries, who are already seeing vast benefits from information technology making development cheaper, and would especially benefit from decreases in the price of sustainable energy generation.

Then there are more subtle reasons to think that technological progress doesn't get the attention it deserves.

First, it works over long time horizons, so it is especially subject to all the kinds of short-termism that plague human decision-making.

Secondly, lost progress isn't visible: if the Internet hadn't been invented, very few would realise what they're missing out on, but try taking it away now and you might well spark a war. This means that stopping technological progress is politically cheap, because likely no one will realise the cost of what you've done.

Finally, making the right decisions about technology is going to decide whether or not the future is good. Debates about technology often become debates about whether we should be pessimistic or optimistic about the impacts of future technology. This is rarely a useful framing, because the only direct impact of technology is to let us make more changes to the world. Technology shouldn't be understood as a force automatically pulling the distribution of future outcomes in a good or bad direction, but as a force that blows up the distribution so that it spans all the way from an engineered super-pandemic that kills off humanity ten years from now to an interstellar civilisation of trillions of happy people that lasts until the stars burn down. Where on this distribution we end up on depends in large part on the decisions we collectively make about technology. So, how about we get those decisions right?

But first, how should we even think about technological progress?

## Modelling technological progress

Some people think that technological progress is stagnating relative to historical trends, and that, for example, we should have flying cars by now. To be able to answer this question, we need some model of what technological progress should be like. I can think of three general ones.

The first one I'll name the Kurzweilian model, after futurist Ray Kurzweil, who's made a big deal about how the intuitive linear model of technological progress is wrong, and history instead shows technological progress is exponential – the larger your technological base, the easier it is to invent new technologies, and hence a graph of anything tech-related should be a hockey-stick curve shooting into the sky.

The second I'll call the fruit tree model, after the metaphor that once the "low-hanging fruit" are picked off, progress gets harder. The strongest case for this model is in science; the physics discoveries you can make by watching apples fall down have (very likely) long since been picked off. However, it's not clear similar arguments should apply to technology. Perhaps we can model inventing a technology as finding a clever way to combine a number of already known parts into a new thing, and hence the number of possible inventions as would be an increasing function of the number of things already invented, since this gives more combinations. For example, even if progress in pure aviation is slow, when we invent new things like lightweight computers we can combine the two to get drones. I haven't seen anyone propose a model to explain why the fruit tree model makes sense for technology in particular.

The third model is that technological change is mostly random. Any particular technological base satisfies the prerequisites for some set of inventions. Once invented, a new technology goes through an S-curve of increasing adoption and development, before reaching widespread adoption and a mature form. Sometimes there are many inventions just within reach, and you get an innovation burst, like the mid-20th century one when television, cars, passenger aircraft, nuclear weapons, birth control pills, and rocketry are all simultaneously going through the rapid improvement and adoption phase. Sometimes there are no plausible big inventions for very long periods of time, for example in medieval times.

Here's an Our World in Data graph (source and interactive version here) showing more-or-less-S-curves for the adoption of a bunch of technologies:

(One can try to imagine an even more general model to unify the three models above, though we're getting to fairly extreme abstraction levels. Nevertheless, for the fun of it: let's model each technology as a set of prerequisite technologies, and assume there's a subset of technology-space that makes up the sensible technologies, and some cost function that describes how hard it is to go from a set of technologies to a given new technology (so infinity if all prerequisites of the new one aren't contained in the known set). Then slow progress would be modelled as the set of sensible ideas and the cost function being such that from any particular set of known technologies, there are only a few sensible ideas with prerequisites only in the known set, and these have high costs. Fast progress is the opposite. In the Kurzweilian model, the subspace of sensible ideas is in some sense uniform, so that the fraction of the $2^{|K|}$ possible prerequisite combinations for a known technology set $K$ that are contained within the sensible set does not go down with the cardinality of $K$, and also we require the cost function to not increase too rapidly as the complexity of the technologies grow. In the fruit tree model, the cost function increases, and possibly the frequency of sensible technologies becomes sparser as you get into the more complex parts of technology-space. In the random model, the cost function has no trend, and a lot of the advancements happen when a "key technology" is discovered that is the last unknown prerequisite for a lot of sensible technologies in technology-space.)

(Question: has anyone drawn up a dependency tree of technologies across many industries (or even one large one), or some other database where each technology is linked to a set of prerequisites? That would be an incredible dataset to explore.)

In Where is my Flying Car?, J. Storrs Hall introduces his own abstraction of a civilisation's technology base that he calls the "technium": imagine some high-dimensional space representing possible technologies, and imagine a blob in this space representing existing technology. This blob expands as our technological base expands, but not uniformly: imagine some gradient in this space representing how hard it is to make progress in a given direction from a particular point, which you can visualise as a "terrain" which the technium has to move along as it expands. Some parts of the terrain are steep: for example, given technology that lets you make economical passenger airplanes moving at near the speed of sound, it takes a lot to progress beyond that because crossing the speed of sound is difficult. Hence the "aviation cliffs" in the image below; the technium is pressing against it, but progress will be slow:

 (Image source: my own slides for an EA Cambridge talk.)

In other cases, there are valleys, where once the technium gets a toehold in it, progress is fast and the boundaries of what's possible gush forwards like a river breaking a dam. The best example is probably computing: figure out how to make transistors smaller and smaller, and suddenly a lot of possibilities open up.

We can visualise the three models above in terms of what we'd expect the terrain to look like as the technium expands further and further:

(Or maybe a better model would be one where the gradient is always be positive, with 0 gradient meaning effortless progress?)

In the Kurzweilian model, the terrain gets easier and easier the further out you go; in the fruit tree it's the opposite; if there is no pattern, then we should expect cliffs and valleys and everything in between, with no predictable trend.

Hall comes out in favour of what I've called the random model, even going as far as to speculate that the valleys might follow a Zipf's law distribution. He concisely summarises the major valleys of the past and future:

"The three main phases of technology that drove the Industrial Revolution were first low-pressure steam engines, then machine tools, and then high-pressure engines enabled by the precision that the machine tools made possible. High-pressure steam had the power-to-weight ratios that allowed for engines in vehicles, notably locomotives and steamships. The three major, interacting, and mutually accelerating technologies in the twenty-first century are likely to be nuclear, nanotech (biotech is the “low-pressure steam” of nanotech), and AI, coming together in a synergy I have taken to calling the Second Atomic Age."

Personally, my views have shifted away from somewhat Kurzweilian ones and towards the random model, with the main factors being that the technological stagnation debate has made me less certain that the historical data fits a Kurzweilian trend, and that since there are no clear answers to whether there is a general pattern, it's sensible to shift the distribution of my beliefs towards the model that doesn't require assuming the truth of a general pattern. However, given some huge valleys that seem to be out there – AI is the obvious one, but also nanotechnology, which might bring physical technology to Moore's law -like growth rates – it is possible that the difference between the Kurzweilian and random model looks largely academic in the next century.

## Is technology stagnating?

Now that we have some idea of how to think about technological progress, we are better placed to answer the question of whether it has stagnated: if the fruit tree model is true we should expect a slowdown, whereas if the extreme Kurzweilian model is true, a single trend line that's not going to break past the top of the figure in the next decade is a failure. Even so, this question is very confusing; economists debate about total factor productivity (a debate I will stay out of), and in general it's hard to know what could have been.

However, it does seem true that compared to the mid-20th century, the post-1970 era has seen breakthroughs in fewer categories of innovation. Consider:

• 1920-1970:

• cars
• television
• antibiotics
• the green revolution
• nuclear power
• passenger aviation
• chemical space travel
• effective birth control
• lasers
• 1970-2020:

• personal computers
• mobile phones
• GPS
• DNA sequencing
• CRISPR
• mRNA vaccines

Of course, it's hard to compare inventions and put them in categories – is lumping everything computing-related as largely the same thing really fair? – but some people are persuaded by such arguments, and a general lack of big breakthroughs in big physical technologies does seem true. (Though might soon change, since the clean energy, biotech, and space industries are making rapid progress.)

Why is this? If we accept the fruit tree model, there's nothing to be explained. If we accept the random one, we can explain it as a fluke of the shape of the idea space terrain that the technium is currently pressing into. To quote Hall again:

"The default [explanation for technological stagnation] seems to have been that the technium has, since the 70s, been expanding across a barren high desert, except for the fertile valley of information technology. I began this investigation believing that to be a likely explanation."

This, I think, is a pretty common view, and is a sensible null hypothesis for the lack of other evidence. We can also imagine variations, like the existence of a huge valley in the form of computing drawing all the talent that would otherwise have gone into pushing the technium forwards in other places. However, Hall rather dramatically concludes that this

"[...] is wrong. As the technium expanded, we have passed many fertile Gardens of Eden, but there has always been an angel with a flaming sword guarding against our access in the name of some religion or social movement, or simply bureaucracies barring entry in the name of safety or, most insanely, not allowing people to make money."

Is this ever actually the case? I think there is a case where a feasible (and economic, environmental, and health-improving) technology has been blocked: nuclear power, as I discuss here. We should therefore amend our model of the technium: not only does it have to contend with the cliffs inherent in the terrain, but sometimes someone comes along and builds a big fat wall on the border, preventing either development, deployment, or both.

In diagram form:

Are there other cases? Yes – GMOs, as I discuss in this review. There have also been some harmful technologies that have been controlled; for example biological and chemical weapons of mass destruction are more-or-less kept under control by two treaties (the Biological Weapons Convention and the Chemical Weapons Convention). However, such cases seem to be the exception, since the overall history is one of technology adoption steamrolling the luddites, from the literal Luddites to George W. Bush's attempts to limit stem cell research.

There are also cases where we put a lot of effort into expanding the technium in a specific direction (German subsidies for solar power are one successful example). We might think of this as adding stairs to make it easier to climb a hill.

How much of the technium's progress (or lack thereof) is determined by the terrain's inherent shape, and how much by the walls and stairs that we slap onto it? I don't know. The examples above show that as a civilisation we sometimes do build important walls in the technium terrain, but arguments like those Hall presents in Where is my Flying Car? are not strong enough to make me update my beliefs to thinking that this is the main factor determining how the technium expands. If I had to make a very rough guess, I'd say that though there is variation based on area (e.g. nuclear and renewable energy have a lot of walls and stairs respectively; computing has neither), overall the inherent terrain has at least several times the effect size on the decadal timescale. The power balance seems heavily dependent on the timescale too – George W. Bush can hold back stem cells for a few years, but imagine the sort of measures it would have taken to delay steam engines for the past few hundred years.

## How should we guide technological progress?

How much should we try to guide technological progress?

A first step might be to look at how good we've been at it in the past, so that we get a reasonable baseline for likely future performance. Our track record is clearly mixed. On one hand, chemical and biological weapons of mass destruction have so far been largely kept under control, though under a rather shoestring system (Toby Ord likes to point out that the Biological Weapons Convention has a smaller budget than an average McDonald's), and subsidies have helped solar and wind to become mature technologies. On the other hand, there are over ten thousand nuclear weapons in the world and they don't seem likely to go away anytime soon (in particular, while New START was recently extended, Russia has a new ICBM coming into service this year and the US is probably going to go ahead with their next-generation ICBM project, almost ensuring that ICBMs – the most strategically volatile nuclear weapons – continue existing for decades more). We've mostly stopped ourselves benefiting from safe and powerful technologies like nuclear power and GMOs for no good reason. More recently, we've failed to allow human challenge trials for covid vaccines, despite massive net benefits (vaccine safety could be confirmed months faster, and the risk to healthy participants is lower than a year at some jobs), an army of volunteers, and broad public support.

Imagine your friend was really into picking stocks, and sure, they once bought some AAPL, but often they've managed to pick the Enrons and Lehman Brothers of the world. Would your advice to them be more like "stay actively involved in trading" or "you're better off investing in an index fund and not making stock-picking decisions"?

Would things be better if we had tried to steer technology less? We'd probably be saving money and the environment (and third-world children) by eating far more genetically engineered food, and air pollution would've claimed millions fewer lives because nuclear power would've done more to displace coal. Then again, we'd probably have significantly less solar power. (Also, depending on what counts as steering technology rather than just reacting to its misuses, we might include the eventual bans on lead in gasoline, DDT, and chloroflourocarbons as major wins.) And maybe without the Biological Weapons Convention becoming effective in 1975, the Cold War arms race would've escalated to developing even more bioweapons than the Soviets already did (for more depth, read this), and an accidental leak might've released a civilisation-ending super-anthrax.

So though we haven't been particularly good at it so far, can we survive without steering technological progress in the future? I made the point above that technology increases the variance of future outcomes, and this very much includes in the negative direction. Maybe hypersonic glide vehicles make the nuclear arms race more unstable and eventually result in war. Maybe technology lets Xi Jinping achieve his dream of permanent dictatorship, and this model turns out to be easily exportable and usable by authoritarians in every country. Maybe we don't solve the AI alignment problem before someone goes ahead and builds one, and the result is straight from Nick Bostrom's nightmares. And what exactly is the stable equilibrium in a world where a 150€ device that Amazon will drone-deliver to anyone in the world within 24 hours can take a genome and print out bacteria and viruses that have it?

This fragility is highlighted in a 2002 paper by Nick Bostrom, who shares the view that the technium can't be reliably held back, at least to the extent that some dangerous technologies might require:

"If a feasible technology has large commercial potential, it is probably impossible to prevent it from being developed. At least in today’s world, with lots of autonomous powers and relatively limited surveillance, and at least with technologies that do not rely on rare materials or large manufacturing plants, it would be exceedingly difficult to make a ban 100% watertight. For some technologies (say, ozone-destroying chemicals), imperfectly enforceable regulation may be all we need. But with other technologies, such as destructive nanobots that self-replicate in the natural environment, even a single breach could be terminal."

The solution is what he calls differential development:

"[We can affect] the rate of development of various technologies and potentially the sequence in which feasible technologies are developed and implemented. Our focus should be on what I want to call differential technological development: trying to retard the implementation of dangerous technologies and accelerate implementation of beneficial technologies, especially those that ameliorate the hazards posed by other technologies." [Emphasis in original]

(See here for more elaboration on this concept and variations.)

For example:

"In the case of nanotechnology, the desirable sequence would be that defense systems are deployed before offensive capabilities become available to many independent powers; for once a secret or a technology is shared by many, it becomes extremely hard to prevent further proliferation. In the case of biotechnology, we should seek to promote research into vaccines, anti-bacterial and anti-viral drugs, protective gear, sensors and diagnostics, and to delay as much as possible the development (and proliferation) of biological warfare agents and their vectors. Developments that advance offense and defense equally are neutral from a security perspective, unless done by countries we identify as responsible, in which case they are advantageous to the extent that they increase our technological superiority over our potential enemies. Such “neutral” developments can also be helpful in reducing the threat from natural hazards and they may of course also have benefits that are not directly related to global security."

One point to emphasise is that the dangerous technology probably can't be held back indefinitely. One day, if humanity continues advancing (as it should), it will be easy to create deadly diseases, build self-replicating nanobots, or spin up a superintelligent computer program in the way that you'd spin up a Heroku server today. The only thing that will save us if the defensive technology (and infrastructure, and institutions) are in place by then. In The Diamond Age, Neal Stephenson imagines a future where there are defensive nanobots in the air and inside people that are constantly on patrol against hostile nanobots. I can't help but think that this is where we're heading. (It's also the strategy our bodies have already adopted to fight off organic nanobots like viruses.)

This is not how we've done technology harm mitigation in the past. Guns are kept in check through regulation, not by everyone wearing body armour. Sufficiently tight rules on, say, what gene sequences you can put into viruses or what you can order your nanotech universal fabricator to produce will almost certainly be part of the solution and go a long way on their own. However, a gun can't spin out of control and end humanity; an engineered virus or self-replicating nanobot might. And as we've seen, our ability to regulate technology isn't perfect, so maybe we should have a backup plan.

The overall picture therefore seems to be that our civilisation's track record at tech regulation is far from perfect, but the future of humanity may soon depend on it. Given this, perhaps it's better that we err on the side of too much regulation – not because it's probably going to be beneficial, but because it's a useful training ground to build up the institutional competence we're going to need to tackle the actually difficult tech choices that are heading our way. Better to mess up regulating Facebook and – critically – learn from it, than to make the wrong choices about AI.

It won't be easy to make the leap from a civilisation that isn't building much nuclear power despite being in the middle of a climate crisis to one that can reliably ensure we survive even when everyone and their dog plays with nanobots. However, an increase in humanity's collective competence at making complex choices about technology is something we desperately need.

RELATED:

## 2021-03-21

### Review: Where is my Flying Car?

Book: Where is my Flying Car?: A Memoir of Future Past, by J. Storrs Hall (2018)

In the 50s and 60s, predictions of the future were filled with big physical technical marvels: spaceships, futuristic cities, and, most symbolically, flying cars. The lack of flying cars has become a cliche, whether as a point about the unpredictability of future technological progress, or a joke about hopeless techno-optimism.

For J. Storrs Hall, flying cars are not a joke. They are a feasible technology, as demonstrated by many historical prototypes that are surprisingly close to futurists' dreams, and practical too: likely to be more expensive than cars, yes, but providing many times more value to owners.

So, where are they?

 Above: not a joke. (Public domain, original here)

The central motivating force behind Where is my Flying Car? is the disconnect between what is physically possible with modern science, and what our society is actually achieving. The immediate objection to such points is to say: "well, of course some engineer can imagine a world where all this fancy technology is somehow economically feasible and widespread, but in the real world everything is more complicated, and once you take these complications into account there's no surprising failure".

Hall's objection is that everything was going fine until 1970 or so.

Many people complain that technological progress has slowed. Flying cars, of course, but also: airliner cruising speeds have stagnated, the space age went on hiatus, cities are still single-level flat designs with traffic, nuclear power stopped replacing fossil fuels, and nanotechnology (in the long run, the most important technology for building anything) is growing slowly. Peter Thiel sums this up by saying "we wanted flying cars, instead we got 140 characters".

It's not just technology. There's an entire website devoted to throwing graphs at you about trends that changed around 1970 (and selling you Bitcoin on the side), and, while a bunch of it is Spurious Correlations material, they include enough important things, like a stagnation in median wages, that it's worth thinking about.

Perhaps the most fundamental indicator is that the energy available per person in the United States was increasing exponentially (a trend Hall names the Henry Adams curve), until, starting around 1970, it just wasn't:

Is this just because the United States is an outlier in energy use statistics? No; other developing countries have plateaued too, with the exception of Iceland and Singapore:

 (Source: Our World in Data, one of the best websites on the internet. You can play around with an interactive version of this chart here.)

Hall tries to estimate what percentage of future predictions in some technical area have come true as a function of the energy intensity of the technology, and finds a strong inverse correlation: in less energy intensive areas (e.g. mobile phones) we've over-achieved relative to futurists' predictions, while the opposite is true with energy intensive big machines (e.g. flying cars). (This is necessarily very subjective, but Hall at least says he did not change any of his estimates after seeing the graph.)

Of course, we have to contrast the stagnation in some areas with enormous advancements during the same time. The most obvious example is computing, something that futurists generally missed. In biotechnology, the price of DNA sequencing has dropped exponentially and in just the past few years we've gotten powerful tools like CRISPR and mRNA vaccines. Meanwhile the average person is now twice as rich as in 1970, and life expectancy has increased by 15 years (and the numbers are not much lower if we restrict our attention just to developed countries).

Perhaps we should be content; maybe Peter Thiel should stop complaining now that we have 280 characters? After all, the problem is not that things are failing, but that they might be improving slower than they could be. That hardly seems like the end of the world. So why should we focus on technological progress? Has it really slowed? And how can we model it? I discuss these questions in another post. In this post, however, I will move straight onto Hall's favourite topic.

## Cool technology

### Flying cars

You might assume the case for flying cars looks something like this:

1. You get to places very fast.
2. Very cool.

However, there's a deeper case to be made for flying cars (or rapid transportation in general), and it starts with the observation that barefoot-walkers in Zambia tend to spend an hour or so a day travelling. Why is this interesting? Because this is the same as the average duration in the United States (of course Hall's other example is the US) or any other society.

Flying cars aren't about the speed – they're about the distance that this speed allows, given universal human preferences for daily travel duration. Cars on the road do about 60 km/h on average for any trip ("you might think that you could do better for a long trip where you can get on the highway and go a long way fast", Hall writes, but "the big highways, on the average, take you out of your way by an amount that is proportional to the distance you are trying to go"). A flying car that goes five times faster lets you travel within twenty-five times the area, potentially opening up a lot of choice.

Hall goes through some calculations about the utilities of different time-to-travel versus distance functions, given empirical results from travel theory, to produce this chart (which I've edited to improve the image quality and convert units) as a summary:

(The overhead time means how long it takes to transition into flying mode, for example if you have to attach wings to it, or drive to an airport to take off.)

Even a fairly lame flying car would easily be three times more valuable than a regular car, mainly by giving you more choice and therefore letting you visit places that you like more.

In terms of what a flying car would actually look like, you have several options. Helicopters are obvious, but they are about ten times the price of cars, mechanically complex (and with very low manufacturing tolerances), and limited by aerodynamics (the advancing blade pushes against the sound barrier, and the retreating one pushes against generating too little lift due to how slowly it moves) to a speed of 250 km/h or so.

Historically, many promising flying car designs that actually flew where autogyros, which generate thrust with a propeller but lift through an unpowered freely-rotating helicopter-like rotor. They generally can't take off vertically, but can land in a very small space.

Another design is a VTOL (vertical take-off and landing) aircraft. Some have been built and used as fighter jets, but they've gained limited use because they're slower and less manoeuvrable than conventional fighters and have less room for weapons. However, Hall notes that one experimental VTOL aircraft in particular – the XV-5 – would "have made one hell of a sports car" and its performance characteristics are recognisable as those of a hypothetical utopian flying car. It flew in 1964, but was cancelled because the Air Force wanted something as fast and manoeuvrable as a fighter jet, rather than "one hell of a sports car".

Of current flying car startups, Hall mentions Terrafugia and AeroMobil, which produce traditional gasoline-powered vehicles (both with fuel economies comparable in litres/km to ordinary cars). There's also Volocopter and EHang, both of which produce electric vehicles with constrained ranges.

Hall divides the roadblocks (or should I say NOTAMs?) for flying cars into four categories.

The first is that flying is harder than driving. To test this idea, Hall learned to fly a plane, and concluded that it is considerably harder, but not insurmountably. Besides, we're not far from self-driving; commercial passenger flights are close to self-piloting already, the existing Volocopter is only "optionally piloted", and the EHang 184 flies itself.

The second is technological. The main challenges here are flying low and slow without stalling (you want to be able to land in small places, at least in emergencies), and reducing noise to manageable levels.

The third is economic. Even though the technology theoretically exists, it may be that we're not yet at a stage where personal flying machines are economically feasible. To some extent this is true; Hall admits that even on the pre-1970 trends in private aircraft ownership, the US private aircraft market would only be something like 30 000 - 40 000 per year (compared to the 2 000 or so that it currently is), about a hundredth of the number of cars sold. The economics means we should expect that the adoption curve is shallow, but not that it's necessarily non-existent.

The final reason is simple: even if you could make a flying car, you wouldn't be allowed to. Everything in aviation is heavily regulated, pushing up costs in a way that, Hall says, leads private pilots to joke about "hundred-dollar burgers". Of course, flying is hard, so you want standards high enough that at the very least you don't have to dodge other people's home-made flying motorbikes as they rain down from the sky, but in Hall's opinion the current balance is wrong.

And it's not just that the balance is wrong, but that the regulations are messed up. For example, making aircraft in the light sports aircraft category would be a great way to experiment with electric flight, but the FAA forbids them from being powered by anything other than a single internal combustion piston engine.

In particular, the FAA "has a deep allergy to people making money with flying machines". If you own a two-seat private aircraft, you can't charge a passenger you take on a flight more than half of the fuel cost, so no air Uber. Until the FAA stopped dragging its feet on drone regulation in 2016, drones were operated under model aircraft rules, and therefore could not be used for anything other than hobby or recreational purposes. Similar rules still apply to ultralights, with one suspicious exception: a candidate for a federal, state, or local election is allowed to pay for a flight.

(And of course, to all these rules it's usually possible to apply for a waiver – so if you're a big company with an army of lawyers, do what you want, but if you're two people in a garage, good luck.)

There's no clear smoking gun of one piece of regulation specifically causing significant harm to flying car innovation. However, the harms of regulation are often a death-by-a-thousand-cuts situation, where a million rules each clip away at what is permissible and each add a small cost. Hall's conclusion is harsh: "It’s clear that if we had had the same planners and regulators in 1910 that we have now, we would never have gotten the family car at all."

One particular effect of flying cars would be to weaken the pull of cities, another topic to which Hall brings a lot of opinions.

### City design

"Designing a city whose transportation infrastructure consists of the flat ground between the boxes is insane."

This is true. Most traffic problems would go away if you could add enough levels. However, "[e]ven the recent flurry of Utopia-building projects are still basically rows of boxes sitting on the dirt plus built-in wifi so the self-driving cars can talk to each other as they sit in automated traffic jams".

As usual, Hall spies some sinister human factors lurking behind the scenes, delaying his visions of techno-utopia:

"There is a perverse incentive for bureaucrats and politicians to force people to interact as much as possible, and indeed to interact in contention, as that increases the opportunities for control and the granting of favors and privileges. This is probably one of the major reasons that our cities have remained flat, one-level no-man’s-lands where pedestrians (and beggars and muggers) and traffic at all scales are forced to compete for the same scarce space in the public sphere, while in the private sphere marvels of engineering have leapt a thousand feet into the sky, providing calm, safe, comfortable environments with free vertical transportation."

This is an interesting idea, and I've read enough Robin Hanson to not discount such perverse explanations immediately, but once again I'm not convinced how important this factor is, and Hall, as usual, is happy to paint only in broad to strokes.

However, he makes a clearly strong point here:

"Densification proponents often point to an apparent paradox: removing a highway which crosses a community often does not increase traffic on the remaining streets, as the kind of hydraulic flow models used by traffic planners had assumed that it would. On the average, when a road is closed, 20% of the traffic it had handled simply vanishes. Traffic is assumed to be a bad thing, so closing (or restricting) roads is seen as beneficial. Well duh. If you closed all the roads, traffic would go to zero. If you cut off everybody’s right foot and forced them to use crutches, you’d get a lot less pedestrian traffic, too."

Hall takes a liberal principle of being strongly in favour of giving people choice, arguing that the goal of city design and transportation infrastructure should be to maximise how far people can travel quickly, rather than trying to ensure that they don't need to travel anywhere other than the set of choices the all-seeing, all-knowing urban designer saw fit to place nearby. Of course, once again flying cars are the best:

"The average American commute to work, one way by car, ranges from 20 minutes to half an hour (the longer times in denser areas).  This gives you a working radius of about 15 miles [= 24 km], or [1800 square kilometres] around home to find a workplace (or around work to find a home). With a fast VTOL flying car, you get a [240-kilometre] radius or [180 thousand square kilometres] of commutable area. Cars, trucks, and highways were clearly one of the major causes of the postwar boom. It isn’t perhaps realized just how much the war on cars contributed to the great stagnation—or how much flying cars could have helped prolong the boom."

### Nuclear power

I discuss nuclear power at length in another post.

### Space travel?

What about the classic example of supposedly stalled innovation – we were on the moon in 1969, and won't return until at least 2024?

"With space travel, there’s a pretty straightforward answer: the Apollo project was a political stunt, albeit a grand and uplifting one; there was no compelling reason to continue going to the moon given the cost of doing so."

The general curve of space progress seems to be over-achievement relative to technological trends in the 60s, followed by stagnation, not because the technology is impossible – we did go to the moon after all – but because it just wasn't economical. Only now, with private space companies like SpaceX and Rocket Lab actually making a business out of taking things to space outside the realm of cosy costs-plus government contracts is innovation starting to pick up again.

(In the past ten years, we've seen the first commercial crewed spacecraft, reuse of rocket stages, the first methane-fuelled rocket engine ever flown, the first full-flow staged-combustion rocket engine ever flown, and the first liquid-fuelled air-launched orbital rocket, just to pick some examples.)

Hall has some further comments about space. First, in this passage he shows an almost-religious deference to trend lines:

"As you can see from the airliner cruising speed trend curve, we shouldn’t have expected to have commercial passenger space travel yet, even if the Great Stagnation hadn’t happened."

I don't think it makes sense to take a trend line for atmospheric flight speeds and use that to estimate when we should have passenger space travel; the physics is completely different, and in particular speeds are very constrained in orbit (you need to go 8 km/s to stay in orbit, and you can't go faster around the Earth without constant thrusting to stop yourself from flying off – something Hall clearly understands, as he explains it more than once).

Secondly, he is of course in favour of everything high-energy and nuclear.

A more controlled alternative to bombing yourself through space is to use a nuclear reactor to heat up propellant in order to expel it out the back of your rocket at high speeds, pushing you forwards. The main limit with these designs is that you can't turn the heat up too much without your reactor blowing up. Hall's favoured solution is a direct fission-to-jet process, where the products of your nuclear reaction go straight out the engine without all this intermediate fussing around with heating the propellant. A reaction that converts a proton and a lithium-7 atom into 2 helium nuclei would give an exhaust velocity of 20 Mm/s (7% of the speed of light), which is insane.

To give some perspective: let's say your design parameters are that you have a 10 ton spacecraft, of which 1 ton can be fuel. With chemical rocket technology, this gives you a little toy with a total ∆V of some 400 m/s, meaning that if you light it up and let it run horizontally along a frictionless train track, it'll break the sound barrier by the time it's out of fuel, but it can't take you from a Earth-to-moon-intercept trajectory to a low lunar orbit even with the most optimal trajectories. With the proton + lithium-7 process Hall describes, your 10% fuel, 10-ton spaceship can accelerate at 1G for two days. If you want to go to Mars, instead of this whole modern business of waiting for the orbital alignment that comes once every 26 months and then doing a 9-month trip along the lowest-energy orbit possible, you can almost literally point your spaceship at Mars, accelerate yourself to a speed of 1 000 km/s over a day (for comparison, the speeds of the inner planets in their orbits are in the tens of kilometres per second range), coast for maybe a day at most, and then decelerate for another day. For most of the trip you get free artificial gravity because your engine is pushing you so hard. This would be technology so powerful even Hall feels compelled to tack on a safety note: "watch out where you point that exhaust jet".

### Nanotechnology!

Imagine if machine pieces could not be made on a scale smaller than a kilometre. Want a gear? Each tooth is a 1km x 1km x 1km cube at least. Want to build something more complicated, say an engine? If you're in a small country, it may well be a necessarily international project, and also better keep it fairly flat or it won't fit within the atmosphere. Want to cut down a single tree? Good luck.

This is roughly the scale at which modern technology operates compared to the atomic scale. Obviously this massively cuts down on what we can do. Having nanotechnology that lets us rearrange atoms on a fine level, instead of relying on astronomically blunt tools and bulk chemical reactions, could put the capabilities of physical technology on the kind of exponential Moore's law curve we've seen in information technology.

There are some problems in the way. As you get to smaller and smaller scales:

• matter stops being continuous and starts being discrete (and therefore for example oil-based lubrication stops working);
• the impact of gravity vanishes but the impact of adhesion increases massively;
• heat dissipation rates increase;
• everything becomes springy and nothing is stiff anymore; and
• hydrogen atoms (other atoms are too heavy) can start doing weird quantum stuff like tunnelling.

Also, how do we even get started? If all we have are extremely blunt tools, how do you make sharp ones?

There are two approaches. The first, the top-down approach, was suggested in a 1959 talk by Richard Feynman, which is credited as introducing the concept of nanotechnology. First, note that we currently have an industrial tool-base at human scales that is, in a sense, self-replicating: it requires human inputs, but we can draw a graph of the dependencies and see that we have tools to make every tool. Now we take this tool-base, and create an analogous one at one-fourth the scale. We also create tools that let us transfer manipulations – the motions of a human engineer's hands, for example – to this smaller-scale version (today we can probably also automate large parts of it, but this isn't crucial). Now we have a tool-base that can produce itself at a smaller scale, and we can repeat the process again and again, making adjustments in line with the above points about how the engineering must change. If each step is one-fourth the previous, 8 iterations will take us from a millimetre-scale industrial base to a tens-of-nanometres-scale one.

The other approach is bottom-up. We already have some ability to manipulate things on the single-digit nanometre scale: the smallest features on today's chips are in this range, we have atomic-scale microscopes that can also manipulate atoms, and of course we're surrounded by massively complicated nanotechnology called organic life that comes with pre-made nano-components. Perhaps these tools let us jump straight to making simple nano-scale machines, and a combination of these simple machines and our nano-manipulation tools lets us eventually build the critical self-sustaining tool-base at the atomic level.

### Weather machines?!

Here's one thing you could do with nanotechnology: make 5 quintillion 1 cm controllable hydrogen balloons with mirrors, release them into the atmosphere, and then set sunlight levels to be whatever you want (without nanotechnology, this might also be doable, but nanotechnology lets you make very thin balloons and therefore removes the need to strip-mine an entire continent for the raw materials).

Hall calls this a weather machine, and it is exactly what it says on the tin, both on a global and local level. He estimates that it would double global GDP by letting regions set optimal temperatures, since "you could make land in lots of places on the earth, such as Northern Canada and Russia, as valuable as California". Of course, this is assuming that we don't care about messing up every natural ecosystem and weather pattern on the planet, but if the machine is powerful enough we might choose to keep the still-wild parts of the world as they are. I don't know if this would work, though; sunlight control alone can do a lot to the weather, but perhaps you'd need something different to avoid, for example, the huge winds from regional temperature differences? However, with a weather machine, the sort of subtle global modifications needed to reverse the roughly 1 watt per square metre increase in incoming solar radiation that anthropogenic emissions have caused would be trivial.

Weather machines are scary, because we're going to need very good institutions before that sort of power can be safely wielded. Hall thinks they're coming by the end of the century, if only because of the military implications: not only could you destroy agriculture wherever you want, but the mirrors could also focus sunlight onto a small spot. You could literally smite your enemies with the power of the sun.

Don't want things in the atmosphere, but still want to control the climate? Then put up sunshades into orbit, incentivising the development of a large-scale orbital launch infrastructure at the same time that we can afterwards use to settle Mars or whatever. As a bonus, put solar panels on your sunshade satellites, and you can generate more power than humanity currently uses.

As always, nothing is too big for Hall. He goes on to speculate about a weather machine Dyson sphere at half the width of the Earth's orbit. Put solar panels on it, and it would generate enormous amounts of power. Use it as a telescope, and you could see a phone lying on the ground on Proxima Centauri b. Or, if the Proxima Centaurians try to invade, you can use it as a weapon and "pour a quarter of the Sun’s power output, i.e. 100 trillion terawatts, into a [15-centimetre] spot that far away,  making outer space safe for democracy."

### Flying cities?!?

And because why the hell not: imagine a 15-kilometre airplane shaped like a manta ray and with a thickness of a kilometre (so the Burj Khalifa fits inside), with room for 10 million people inside. It takes 200 GW of power to stay flying – equivalent to 4 000 Boeing 747s – which could be provided by a line of nuclear power plants every 100 metres or so running along the back. This sounds like a lot, but Hall helpfully points out the reactors would only be 0.01% of the internal volume, so you could still cluster Burj Khalifas inside to your heart's content, and the energy consumption comes out to only 20 kW per person, about where we'd be today if energy use had continued growing on pre-1970s trends.

If you don't want to go to space but still want to leave the Earth untouched, this is one solution, as long as you don't mind a lot of very confused birds.

## Technology is possible, but has risks

I worry that Where is my Flying Car? easily leaves the impression that everything Hall talks about is part of some uniform techno-wonderland, which, depending on your prior about technological progress, is somewhere between certainly going to happen or permanently relegated to the dreams of mad scientists. Hall does not work to dispel this impression: he goes back and forth between talking about how practical flying cars are and exotic nuclear spacecraft, or between reasonable ideas about traffic layout in cities and far-off speculation about city-sized airplanes. Credible world-changing technologies like nanotechnology easily seem like just another crazy thought Hall sketched out on the back of the envelope and could not stop being enthusiastic about.

So should we take Hall's more grounded speculation seriously and ignore the nano-nuclear-space-megapolises? I think this would be the wrong takeaway. First, I'm not sure Hall's crazy speculation is crazy enough to capture possible future weirdness within it; he restricts himself mainly to physical technologies, and thus leaves out potentially even weirder things like a move to virtual reality or the creation of superhuman intelligence (whether AI or augmented humans).

Second, Hall does have a consistent and in some way realist perspective: if you look at the world – not at the institutions humans have built, or whatever our current tech toolbox contains, but at the physical laws and particles at our disposal – what do you come up with?

After all, our world is ultimately not one of institutions and people and their tools. The "strata" go deeper, until you hit the bedrock of fundamental physics. We spend most of our time thinking about the upper layers, where the underlying physics is abstracted out and the particles partitioned into things like people and countries and knowledge. This is for good reason, because most of the time this is the perspective that lets you best think about things important to people. Occasionally, however, it's worth taking a less parochial perspective by looking right down to the bedrock, and remembering that anything that can be built on that is possible, and something we may one day deal with.

This perspective should also make clear another fact. The things we care about (e.g. people) exist many layers of abstraction up from the fundamental physics, and are therefore fragile, since they depend on the correct configuration of all levels below. If your physical environment becomes inhospitable, or an engineered virus prevents your cells from carrying out their function, the abstraction of you as a human with thoughts and feelings will crash, just like a program crashes if you fry the circuits of the computer it runs on.

So there are risks, new ones will appear as we get better at configuring physics, and stopping civilisation from accidentally destroying itself with some new technology is not something we're automatically guaranteed to succeed at.

Hall does not seem to recognise this. Despite all his talk about nanotechnology, the grey goo scenario of self-replicating nanobots going out of control and killing everyone doesn't get a mention. As far as I'm aware, there's no strong theoretical reason for this to be impossible – nanobots good at configuring carbon/oxygen/hydrogen atoms are a very reasonable sort of nanobot, and I can't help but noticing that my body is mainly carbon, oxygen, and hydrogen atoms. "What do you replace oil lubrication with for your atomic scale machine parts" is a worthwhile question, as Hall notes, but I'd like to add that so is the problem of not killing everyone.

Hall does mention the problem of AI safety:

"The latest horror-industry trope is right out of science fiction [...]. People are trying to gin up worries that an AI will become more intelligent than people and thus be able to take over the world, with visions of Terminator dancing through their heads. Perhaps they should instead worry about what we have already done: build a huge, impenetrably opaque very stupid AI in the form of the administrative state, and bow down to it and serve it as if it were some god."

What's this whole thing with arguments of the form "people worry about AI, but the real AI is X", where X is whatever institution the author dislikes? Here's another example from a different political perspective (by sci-fi author Ted Chiang, whose fiction I enjoy). I don't think this is a useless perspective – there is an analogy between institutions that fail because their design optimises for the wrong thing, and the more general idea of powerful agents accidentally designed to optimise for the wrong thing – but at the end of the day, surprise surprise, the real AI is a very intelligent computer program.

Hall also mentions he "spent an entire book (Beyond AI) arguing that if we can make robots smarter than we are, it will be a simple task to make them morally superior as well." This sounds overconfident – morality is complicated, after all – but I haven't read it.

As for climate change, Hall acknowledges the problem but justifies largely dismissing it by citing “[t]he actual published estimates for the IPCC’s worst case scenario, RCP8.5, [which] are for a reduction in GDP of between 1% and 3%". This is true ... if you only consider the United States! (The EU is in the same range but the global estimates range up to 10%, because of a disproportionate effect on poor tropical countries.) As the authors of that very report also note, these numbers don't take into account non-market losses. If Hall wants to make an argument for techno-optimistic capitalism, he should consider taking more care to distinguish himself from the strawman version.

## It's not the technology, stupid!

Hall does not think that we'd have all the technologies mentioned above if only technological progress had not "stagnated". The things he expects could've happened by now given past trends are:

• The technological feasibility of flying cars would be demonstrated and sales would be on the rise; Hall goes as far as to estimate the private airplane market in the US could have been selling 30k-40k planes per year (a fairly tight confidence interval for something this uncertain); compare with the actual US market today, which sells around 16 million cars and a few thousand private aircraft per year.
• Demonstrated examples of multi-level cities and floating cities.
• Chemical spacecraft technology would be about where they are now, but some chance that government funding would have resulted in Project Orion-style nuclear launch vehicles.
• Nanotechnology: basic things like ammonia fuel cells might exist, but not fancier things like cell repair machines or universal fabricators.
• Nuclear power would generate almost all electricity, and hence there would be a lot less CO2 in the atmosphere (this study estimates 174 billion fewer tons of CO2 had reasonable nuclear trends continued, but Hall optimistically gives the number as 500 billion tons).
• AI and computers at the same level as today.
• A small probability that something unexpected along the lines of cold fusion would have turned out to work and been commercialised.
• A household income several times larger than today.

So what went wrong? Hall argues:

"The faith in technology reflected in Golden Age SF and Space Age America wasn’t misplaced. What they got wrong was faith in our culture and bureaucratic arrangements."

He gives two broad categories of reasons: concrete regulations, and a more general cultural shift from hard technical progress to worrying and signalling.

### Regulation ruins everything?

Hall does not like regulation. He estimates that had regulation not grown as it did after 1970, the increased GDP growth might have been enough to make household incomes 1.5 to 2 times higher than they are today in the US. I can find some studies saying similar things – here is one claiming 0.8% lower GDP growth per year since 1980 due to regulation, which would imply today's economy would be about 1.3 times larger had this drag on growth existed. As far as I can tell, these estimates also don't take into account the benefits of regulation, which are sometimes massive (e.g. banning lead in gasoline). However, I think most people agree that regardless of how much regulation there should be, it could be a lot smarter.

Hall's clearest case for regulation having a big negative impact on an industry is private aviation in the United States, which crashed around 1980 after more stringent regulations were introduced. The number of airplane shipments per year dropped something like six-fold and never recovered.

A much bigger example is nuclear power, which I will discuss in an upcoming post, and which Hall also has plenty to say about.

Strangely, Hall misses perhaps the most obvious case in modern times: GMOs pointlessly being almost regulated out of existence, a story told well in Mark Lynas' Seeds of Science (my review here). Perhaps this is because of Hall's focus on hard sciences, or his America-centrism (GMO regulation is worse in the EU than in the United States).

And speaking of America-centrism, the biggest question I had is why even if the US is bad at regulation, no country decides to do better and become the flying car capital of the world. Perhaps good regulation is hard enough that no one gets it right? Hall makes no mention of this question, though.

He does, however, throw plenty of shades on anything involving centralisation. For example:

"Unfortunately, the impulse of the Progressive Era reformers, following the visions of [H. G.] Wells (and others) of a “Scientific Socialism,” was to centralize and unify, because that led to visible forms of efficiency. They didn’t realize that the competition they decried as inefficient, whether between firms or states, was the discovery procedure, the dynamic of evolution, the genetic algorithm that is the actual mainspring of innovation and progress."

He brings some interesting facts to the table. For example, an OECD survey found a 0.26 correlation between private spending on research & development and economic growth, but a -0.37 between public R&D and growth. Here's Hall's once again somewhat dramatic explanation:

“Centralized funding of an intellectual elite makes it easier for cadres, cliques, and the politically skilled to gain control of a field, and they by their nature are resistant to new, outside, non-Ptolemaic ideas. The ivory tower has a moat full of crocodiles.”

He backs this up with his personal experiences of US government spending on nanotechnology lead to a flurry of scientists trying to claim that their work counted as nanotechnology (up to and including medieval stained glass windows) as well as trying to discredit anything that actually was nanotechnology, to make sure that the nanotechnologists wouldn't steal more federal funding in the future.

Studies, not surprisingly, find that the issue is more complicated (see for example here, which includes a mention of the specific survey Hall references).

Hall also includes a graph of economic growth vs the Fraser Institute's economic freedom score in the United States. I've created my own version below, including some more information than Hall does:

In general, it seems sensible to expect economic freedom to increase GDP: the more a person's economic choices are limited, the more likely the limitations are to prevent them from taking the optimal action (the main counterexample being if optimal actions for an individual create negative externalities for society). We can also see that this is empirically the case – developed countries tend to have high economic freedom. However, in using this graph as clear evidence, I think Hall is once again trying to make too clear a case on the basis of one correlation.

Effective decentralised systems, whether markets or democracy, are always prone to attack by people who claim that things would be better if only we let them make the rules. Maybe it takes something of Hall's engineer mindset to resist this impulse and see the value of bloodless systems and of general design principles like feedback and competition. (And perhaps Hall should apply this mindset more when evaluating the strength of evidence for his economic ideas.)

As for what the future of societal structure looks like, Hall surprisingly manages to avoid proposing flying-car-ocracy:

""[It] may well be possible to design a better machine for social and economic control than the natural marketplace. But that will not be done by failing to understand how it works, or by adopting the simplistic, feedback-free methods of 1960s AI programs. And if ever it is done, it will be engineers, not politicians, who do it."

He goes further:

"As a futurist, I will go out on a limb and make this prediction: when someone invents a method of turning a Nicaragua into a Norway, extracting only a 1% profit from the improvement, they will become rich beyond the dreams of avarice and the world will become a much better, happier, place. Wise incorruptible robots may have something to do with it."

### Risk perception and signalling

Hall's second reason for us not living up to expectations for technological progress is cultural. He starts with the idea of risk homeostasis in psychology: everyone has some tolerance for risk, and will seek to be safer when they perceive current risk to be higher, and take more risks when they perceive current risk to be lower. In developed countries, risks are of course ridiculously low compared to historical levels, so most people feel safer than ever. Some start skydiving in response, but Hall suggests there's another effect that happens when an entire society finds itself living below their risk tolerance:

"One obvious way [to increase perceived risk] is simply to start believing scare stories, from Corvairs to DDT to nuclear power to climate change. In other words, the Aquarian Eloi became phobic about everything specifically because we were actually safer, and needed something to worry about."

I know what you're thinking – what the hell are "Aquarian Eloi"? Hall likes to come up with his own terms for things, and in this case he is making a reference to H. G. Wells' The Time Machine, in which descendants of humanity live out idle and dissolute lives (modelled on England's idle rich of the time), in order to label what he claims is the modern zeitgeist. Yes, this book is weird at times.

Another cultural idea he touches on is increased virtue signalling. Using the idea of Maslow's hierarchy of needs, he explains that as more and more of the population is materially well-off, more people invest more effort into self-actualisation. Some of this is productive, but, humans being humans, a lot of this effort goes into trying to signal how virtuous you are. Of course, there's nothing inherently wrong with that, as long as your virtue signalling isn't preventing other people climbing up from lower levels of Maslow's hierarchy – or, Hall would probably add, from building those flying cars.

### Environmentalism vs Greenism

A particular sub-case of cultural change that Hall has a lot to say about is the "Green religion", something he distinguishes (though sometimes with not enough care) from perfectly reasonable desires "to live in a clean, healthy environment and enjoy the natural world".

This ideological, fear-driven and generally anti-science faction within the environmentalist movement is much the same thing as what Steven Pinker calls "Greenism", which I talked about in my review of Enlightenment Now (search for "Greenism") and also features in my review of Mark Lynas' Seeds of Science (search for "torpedoes"). Unlike Lynas or even Pinker, Hall does not hold back when it comes to criticising this particular strand of environmentalism. He explains it as an outgrowth of the risk-averseness and virtue signalling trends described above. The "Green religion", he claims, is now the "default religion of western civilization, especially in academic circles", and "has developed into an apocalyptic nature cult". To explain its resistance to progress and improving the human condition, he writes:

"It seems likely that the fundamentalist Greens started with the notion that anything human was bad, and ran with the implication that anything that was good for humans was bad. In particular, anything that empowered ordinary people in their multitudes threatened the sanctity of the untouched Earth. The Green catechism seems lifted out of classic Romantic-era horror novels. Any science, any engineering, the “acquirement of knowledge,” can only lead to “destruction and infallible misery.” We must not aspire to become greater than our nature."

There are troubling tendencies in ideological Greenism (as there is with anything ideological), but I think "apocalyptic nature cult" takes it too far, and as a substitute religion for the west, it has some formidable competitors. Hall is right to point out the tension between improving human welfare and Greenist desires to limit humans, but I'd bet that the driving factor isn't direct disdain for humans, but rather the sort of sacrificial attitudes that are common in humans (consider the people who went around whipping themselves during the Black Death to try to atone for whatever God was punishing them for). Probably there's some part of human psychology or our cultural heritage that makes it easy to jump to sacrifice, disparaging ourselves (or even all of humanity), and repentance as the answer to any problem. While this a nobly selfless approach, it's just less effective than, and sometimes in opposition to, actually building things: developing new technologies, building clean power plants, and so on.

Hall also goes too far in letting the Greenists tar his view of the entire environmentalist movement. Not only is climate change a more important problem than the 1-3% estimated GDP loss for the US suggests, but you'd think that the sort of big technical innovation that is happening with clean tech would be exactly the sort of progress Hall would be rooting for.

Hall does have an environmentalist proposal, and of course it involves flying cars:

"The two leading human causes of habitat destruction are agriculture and highways—the latter not so much by the land they take up, but by fragmenting ecosystems.  One would think that Greens would be particularly keen for nuclear power, the most efficient, concentrated, high-tech factory farms, and for ... flying cars. "

[Ellipsis in original]

### Energy matters!

Despite being partly blinded by his excessive anti-Greenism, there is one especially important correction to some strands of environmentalist thinking that Hall makes well: cheap energy really matters and we need more of it (and energy efficiency won't save the day).

Above, I used the stagnation in energy use per capita as an example of things going wrong. This may have raised some eyebrows; isn't it good that we're not consuming more and more energy? Don't we want to reduce our energy consumption for the sake of the environment?

First, it is obviously true that we need to reduce the environmental impact of energy generation. Decoupling GDP growth from CO2 emissions is one of the great achievements of western countries over the past decades, and we need to massively accelerate this trend.

However, our goal, if we're liberal humanists, should be to give people choices and let them lead happy lives (while applying the same considerations to any sentient non-human beings, and ideally not wrecking irreplaceable ecosystems). In our universe, this means energy. Improvements in the quality of life over history are, to a large extent, improvements in the amount of energy each person has access to. This is very true:

“Poverty is ameliorated by cheap energy. Bill Gates, nowadays perhaps the world’s leading philanthropist, puts it, “If you could pick just one thing to lower the price of—to reduce poverty—by far you would pick energy.”"

Even in the United States, "[e]nergy poverty is estimated to kill roughly 28,000 people annually in the US from cold alone, a toll that falls almost entirely on the poor".

Climate change cannot be solved by reducing energy consumption, because there are six billion people in the world who have not reached western living standards and who should be brought up to them as quickly as possible. This will take energy. What we need is to simultaneously massively increase the amount of energy that humanity uses, while also switching over to clean energy. If you think only one of these is enough, you have either failed to understand the gravity of the world's poverty situation or the gravity of its environmental one.

(Energy efficiency matters, because all else being equal, it reduces operating costs. It is near-useless for solving emissions problems, however, because the more efficiently we can use energy, the more of it we will use. Hall illustrates this with a thought experiment of a farmer who uses a truck to carry one crate of tomatoes at a time from their farm to a customer, and whose only expense is fuel for the truck. Double its fuel efficiency, and it's economical to drive twice as far, and hence service four times as many customers (assuming customer number is proportional to reachable area), plus each trip is twice as long on average. The net result is that the 2x increase in efficiency leads to 8x more kilometres driven and hence 4x higher fuel consumption. The general case is called Jevons paradox.)

So yes, we need energy, most urgently in developing countries, but the more development and deployment of new energy sources there is, the cheaper they will be for everyone – consider Germany's highly successful subsidies for solar power – so developed countries have a role to play as well. (Also, are we sure there would be no human benefits to turning the plateauing in developed country energy use back into an increase?)

You'd think this is obvious. Unfortunately it isn't. In a section titled ""AAUGHH!!", Hall presents these quotes:

“The prospect of cheap fusion energy is the worst thing that could happen to the planet. —Jeremy Rifkin

Giving society cheap, abundant energy would be the equivalent of giving an idiot child a machine gun. —Paul Ehrlich

It would be little short of disastrous for us to discover a source of clean, cheap, abundant energy, because of what we might do with it. —Amory Lovins”

They are what leads Hall to say, perhaps with too much pessimism:

"Should [a powerful new form of clean energy] prove actually usable on a large scale, they would be attacked just as viciously as fracking for natural gas, which would cut CO2 emissions in half, and nuclear power, which would eliminate them entirely, have been."

It is good to give people the choice to do what they want, and therefore good to give them as much energy as possible to play with, whether they want it to power the construction of their dream city or their flying car trips to Australia (I do draw the line at Death Stars, though).

Right now we're limited by the wealth of our societies, limiting us to about 10 kW/capita in developed countries, and by the unacceptable externalities of our polluting technology. The right goal isn't to enforce limits on what people can do (except indirectly through the likes of taxes and regulation to correct externalities), but to bring about a world where these limits are higher.

If energy is expensive, people are cheap – lives and experiences are lost for want of a few watts. This is the world we have been gradually dragging ourselves out of since the industrial revolution, and progress should continue. Energy should be cheap, and people should be dear.

## Don't panic; build

Where is my Flying Car? is a weird book.

First of all, I'm not sure if it has a structure. Hall will talk about flying cars, zoom off to something completely different until you think he's said all he has to say on them, and just when you least expect it: more flying cars. The same pattern of presentation repeats with other topics. Also, sections begin and sometimes end with a long selection of quotes, including no less than three from Shakespeare.

Second, the ideas. There are the hundred speculative examples of crazy (big, physical) future technologies, the many often half-baked economic/political arguments, the unstated but unmissable America-centrism, and witty rants that wander the border between insightful social critique and intellectualised versions of stereotypical boomer complaints about modern culture.

Also, the cover is this:

 Above: ... a joke?

However, I think overall there's a coherent and valuable perspective here. First, Hall is against pointless pessimism. He makes this point most clearly when talking about dystopian fiction, but I think it generalises:

"Dystopia used to be a fiction of resistance; it’s become a fiction of submission, the fiction of an untrusting, lonely, and sullen twenty-first century, the fiction of fake news and infowars, the fiction of helplessness and hopelessness. It cannot imagine a better future, and it doesn’t ask anyone to bother to make one. It nurses grievances and indulges resentments; it doesn’t call for courage; it finds that cowardice suffices. Its only admonition is: Despair more."

Hall's answer to this pessimism is to point out ten billion cool tech things that we could do one day. He veers too much to the techno-optimistic side by not acknowledging any risks, but overall this is an important message. Visions of the future are often dominated by the negatives: no war, no poverty, no death. Someone needs to fill in the positives, and while Hall focuses more on the "what" of it than the "how does it help humans" part, I think a hopeful look at future technologies is a good start.

In addition to being against pessimism about human capabilities, Hall also takes, at least implicitly, a liberal stand by being against pessimism about humans. His answer to "what should we do?" is to give people choice: let them travel far and easily, let them live where they want, let them command vast amounts of energy.

Hall also identifies two ways to keep a civilisation on track in terms of making technological progress and not getting consumed by signalling and politics: growing, and having a frontier.

On the topic of growth, he makes basically the same point as my post on growth and civilisation:

"One of the really towering intellectual achievements of the 20th Century, ranking with relativity, quantum mechanics, the molecular biology of life, and computing and information theory, was understanding the origins of morality in evolutionary game theory. The details are worth many books in themselves, but the salient point for our purposes is that the evolutionary pressures to what we consider moral behavior arise only in non-zero-sum interactions. In a dynamic, growing society, people can interact cooperatively and both come out ahead. In a static no-growth society, pressures toward morality and cooperation vanish; you can only improve your situation by taking from someone else. The zero-sum society is a recipe for evil."

Secondly, the idea of a frontier: something outside your culture that your society presses against (ideally nature, but I think this would also apply to another competing society). This is needed because"[w]ithout an external challenge, we degenerate into squabbling [and] self-deceiving".

"But on the frontier, where a majority of one’s efforts are not in competition with others but directly against nature, self-deception is considerably less valuable. A culture with a substantial frontier is one with at least a countervailing force against the cancerous overgrowth of largely virtue-signalling, cost-diseased institutions."

Frontiers often relate to energy-intensive technologies:

"High-power technologies promote an active frontier, be it the oceans or outer space. Frontiers in turn suppress self-deception and virtue signalling in the major institutions of society, with its resultant cost disease. We have been caught to some extent in a self-reinforcing trap, as the lack of frontiers foster those pathologies, which limit what our society can do, including exploring frontiers. But by the same token we should also get positive feedback by going in in the opposite direction, opening new frontiers and pitting our efforts against nature."

Finally, Hall's book is a reminder that an important measure to judge a civilisation against is its capacity to do physical things. Even if the bulk of progress and value is now coming from less material things, like information technology or designing ever fairer and more effective institutions, there are important problems – covid vaccinations, solving climate change, and building infrastructure, for example – that depend heavily on our ability to actually go out and move atoms in the real world. Let's make sure we continue to get better at that, whether or not it leads to flying cars.

RELATED: