Boris Smus

interaction engineering

Grand Transitions by Vaclav Smil

Up front approach is interesting. Having worked with Jay Forester at MIT on early modeling, Smil is not a big believer in models.

Instead of simple models Smil takes a more contingent, event based approach to the topic. Some things progress in an evolutionary way. But others are dependent on human action and sudden unexpected discontinuities. Even the title “Grand Transitions” harkens to a punctuated equilibrium view of progress.

Smil alludes to the Expected vs unexpected inventions framing I first heard from Clarke (see Profiles of the Future by Arthur Clarke). He thinks that information scaling inventions like the printing press and the internet tend are especially likely to have unexpected consequences and break existing models.

S curves everywhere

Reminiscent of Perez in her Installation vs deployment phases of innovation (Carlotta Perez), Smil sees S-curves everywhere and warns about over-extrapolating exponential growth. Reminds me that Forecasting s-curves is hard.

Human ➡ animal ➡ machine power: draft horses peak in 1910 at 20 million draft horses in the US and are then obsoleted by mechanical tractors, but this transition takes decades. As they grow, populations of draft horses and tractors form S curves:

Transatlantic travel only became possible in the 15th century. Then sailing frigates took about a month to cross. Steam powered ships that would eventually take an order of magnitude less time (Lucitania in 4 days). This transition took a century around 1830-1930. Then the transition to jet air travel, yielded a further reduction of 1.5 orders of magnitude (5h). This was a much faster transition of just a decade or so in the 1950s.

But not all “epochal transitions” pan out as S-curves. Some exponential growth turns out to be a false start in retrospect.

Early electric cars were the first and most promising ground transport in the late 19th century. The first car to reach 100 km/h was electric. Edison bet big on them but gas internal combustion won instead. Smil doesn’t elaborate as to why but I suspect energy density has a big role to play (more on this later). It isn’t until now that the epochal transition to EVs is underway. But predictions are such that we won’t have majority electric on the roads until at least 2040.

Nuclear energy generation is another example of a false start that looks even more bleak. Peaking at 17% of global energy output, current projections are just 4% of global energy by 2040. Culprits appear to be going over budget and safety. Also see Nuclear power is hampered by regulation (ALARA).

Demographic transitions

The developed world is characterized by a shift in equilibrium from high fertility and high mortality to a new one, balancing low fertility and low mortality.

Demographic dividend refers to the benefit a polity experiences when the majority of its citizens are of working age. The ratio of working population (20-65) to their dependents (young and old) is an interesting metric to track. A society with high mortality and high fertility has an young average age, and many young dependents. As mortality and fertility decrease, the average citizen is of working age and the society reaches a stage of development that pays a high demographic dividend because many workers support few dependents. In next equilibrium, the average population becomes old and once again there are many dependents.

Median ages vary hugely per country. The Japanese median age is 47.3, while the Guatemalan median age is 22.1. The demographic dividend will be greatly advantageous to young countries (see chart).

Smil is pretty pessimistic about lifespan extension because he sees it as an S-curve that tapers out at around 80 years old. Increased societal longevity should definitely come with an older retirement age!

Japan leads the way here and is projected to be depopulating at a huge rate. 120 M now. 100 M by 2050. 88 M stable state. Villages are shrinking.

Urban vs rural footprints: Smil criticizes cities for being terrible polluters. This directly contradicts the common wisdom that cities and dense living is far more efficient per-capita. Isn’t it actually quite good to be in a city because of the benefits of proximity?

Smil's argument is that cities are actually worse per capita from an environment perspective. A villager that moves to town may have been used to one light per room but now has a bigger TV. Another urban downside is that heat islands are 3-8 degrees higher in cities.

Which one is right?

Humans win mammal population density: 50k humans / square km density in parts of Manila is equivalent to 2kg of human biomass per square m. This is than for any mammal.

Economic transitions

Although agriculture is only 0.7% of the GDP, it is underrated: “Let those economists live off the main sector of GDP: financial services,” quips Smil.

Agriculture improvements: Huge yield gains. Lots of automation in food production, clear when you consider output in terms of time per weight of output. (Interesting idea: use agriculture as a terminus for Asimov's Tech Tree)

  • 1800: 7 minutes per kilogram of wheat. Tools: two oxen, a wood plow, brush harrows, sickles to cut, flails to thresh.
  • 1900: 25 seconds per kilogram of wheat, an 18x improvement. heavy horses, a steel gang plow led by a team, iron tooth harrows, seed drill, horse drawn combine for harvesting and threshing.
  • 2000: 6 seconds per kilogram, a 4x improvement. Large tractors are equivalent to 500 horses, high capacity combines.

Farming trends:

  • Specialization: 5 crops per farm in 1900 -> 1.1 crops per farm in 2000.
  • Centralization: fewer but larger farms.

Energy subsidies: Modern agriculture requires a lot of additional energy beyond just the natural sunlight, in the form of fertilizer, irrigation, fuel for farm equipment. 4% of global energy goes to these energy subsidies. Without them, we couldn’t feed the current population of 7.5B.

Animals are especially energy-intensive: Based on this line of thinking, animals cost a lot more based on just the energy required to produce the feed. Animals also require structures and maintenance of those structures. And there are additional costs around transportation.

History of crop rotation technology

Fertilizers and side effects: Guano was used as early fertilizer. By 1824, there was an active Peruvian import business. Nitrate was the first inorganic fertilizer. Use of Fertilizers meant crops grow better and are more attractive to pests. Thus, insecticides.

Feeding draft animals: There is a complex relationship between draft animals, and the feed they require which detracts from crop yields. More than half of all food production is animal feed (!). Smil suggests that animals could be far more sustainable if they were allowed to graze naturally.

(There's a nice opportunity to explain these things using feedback loops, along the lines of Horses, peasant mobility and urbanization. Just like in the middle ages, draft animals and farm mobility in general cuts across all these aspects: planting, harvesting with combines, and transportation.

Animal mass to feed ratios: A metric to measure the efficiency of animal protein production. Input is weight of feed, output is the animal weight that can be consumed.

  • Chickens: 3:1 (3 kg and feed produces 1 kg of chicken)
  • Pork: 9:1
  • Beef: 25:1

For aquaculture it’s 3:1 but varies with species. For example, it's 1.8 for farmed Atlantic Salmon.

Dietary patterns: Two major attractors for dietary patterns as societies become more affluent.

  1. Western-style 3200 kcal per day and 30% of calories from animal products.
  2. Mediterranean/Asian style has fewer calories per day (2400 kcal) and less reliance on animal products.

Remarkable stats in Japan. Rice intake (by weight) went from 60-70 percent in 1950 to just 20% today. Now the Japanese diet is more dairy than rice (by calories) and this shift in just 50 years post WW2.

Smil is genuinely impressed with the long way we’ve come on ending malnutrition and famines, though a combination of increased farming efficiency and science (see List of micronutrients and deficiencies in humans). But he's miffed that this truly amazing story gets so little airtime in terms of “great innovations to celebrate” compared to smartphones and Steve Jobs and as he disparagingly puts it “putting new cardboard on electronics manufactured in China”.

This reminds me a lot of this idea from What can a technologist do to accelerate electrification:

The inconveniences of daily life are not the significant problems. The world that scrolls past you on Twitter is not the real world.

Energy transitions

The usual energy progression is this:

  1. Wood
  2. Coal
  3. Oil
  4. Gas

Smil tracks the dates at which each country transitioned from mostly X to mostly Y. Transition years vary widely by country. England moved to coal very early and this catalyzed the Industrial Revolution.

Some countries and regions skipped whole steps, for example going straight from wood burning to oil and never ramping up and coal industry, or jumping directly to hydroelectricity.

Transportation: There's an impressive increase in shipping container payload: from 100 twenty foot equivalent unit (TEU) in 1950 to 23,000 TEU in 2020.

Kerosene vs. diesel vs. gasoline, etc: Oil (aka Petroleum) is generally found as a yellowish-brown liquid in geological formations. It can then be refined into a variety of fuels. Kerosene (aka paraffin) is the basis for airplane fuel. Diesel fuel has many applications from small engines powering a private car to large marine engines. There are many other kinds of petroleum based fuel types, including Heavy Fuel Oil (HFO) which powers some of the largest container ships.

Air travel: Kilometers traveled per passenger is a useful metric for quantifying air travel. It's hard to predict where this S-curve will level out.

Horses vs oxen: Smil estimate of horse power vs ox power is surprising. 7x human for horse and only 2.5x human for ox?! This is surprising, since I read elsewhere (Medieval Technology and Social Change by Lynn White) that horses were quite weak compared to oxen, although this was before selective breeding made them stronger, and before horse collars.

Power-to-weight ratios: Turbofans have the highest watts per kg density of any machine. (How does my e-foil motor compare? Back of napkin calculation the 65161 has max output 6 kW and weight 3 kg.)

Electricity is amazing: Smil writes a real ode to electricity. It's a marvel that powers all of modernity: lossless to deliver from station to home. Completely silent and non polluting at conversion point. Safe. Easy to convert to motion heat and light. The world saw extremely rapid electrification of urban and rural households. From start to finish, the S-curve took only 75 years. Energy transitions take decades, even if highly desirable.

Lumens per watt: A metric for measuring light efficiency. Incandescent to CFL to LED. It's plummeted over the last two centuries.

Cooking range transitions: early ranges were electric, then switched to gas and are now returning to electric induction stoves. Has this transition happened?

So many air conditioners: Wild that us a conditioner penetration is 95% in the US. I double checked this, Google says 90%.

Efficiency in transport: Not covered directly but I wondered how to compare various modes of transportation. The result is my note on Efficiency of transportation modes.

Efficiency gains without drastic changes: We tend to underrate and under report incremental improvements as key components of an invention. The first steam engine in the 18th century had a 1% efficiency. Watt's improvements doubled it to just 2%. But then subsequent unnamed inventions made efficiency rise to 8%.

Modern gasoline engines are substantially more efficient (usually 30-35% efficiency). Diesel is a bit better, around 45% efficient.

Economic transitions

There is no clear marker for completed economic transition. For demographics, the transition happens when the birth rate plummeted below replacement. For energy, a transition happened when a country fully switches from plant based fuels.

Growth is slowing: Also, GDP over time curves appear as S curves with inflection points in the recent past for most developed countries. This means growth is slowing. However, it's unclear if this GDP stat can be trusted.

One economic transition model:

  1. Primary markets: focus on raw materials.
  2. Secondary markets: focus on refining the materials, creating goods from them.
  3. Tertiary markets: everything else that meets consumer demands. Includes services.

Primitive societies are heavily skewed towards primary markets. As societies progress, they tend to shift to tertiary markets, focusing on services. Here's the graph for the US:

Another economic transition involves more people traveling further. This includes for work commutes as well as travel for leisure. Since 1800, people travel an order of magnitudes more for work than before.

Electronics are surprisingly energy intensive: Total energy costs to manufacture all electronics (computers, phones etc) was 1 exajoule (EJ). Compared to all cars at 7EJ. This despite the fact that all cars manufactured weigh 100x as much. Why? Partly due to planned obsolescence and rapid improvement of electronics. (I really should strive to Buy it for life with these electronics devices as much as possible. Optimize for longevity. Including my current iPhone. It’s just wasteful otherwise.)

Climate transitions

Interesting foray into anthropogenic climate change from pre-modern humans:

  1. Impact of runaway fires from cooking accidents
  2. Extinction of mega fauna due to migrating hunters (and climate changes)
  3. Impact of early farming on soil PH and other indicators
  4. Deforestation

Forest density on ice-free land: In terms of tree cover, the metric is mainly forests as measured by sufficient tree density on ice free land. So that means you can take a country and divide its total land by total ice-free land by the percentage of that ice-free land that has forest on it. By this metric most developed world countries have increasing tree cover. And furthermore as climate reduces the amount of ice covered land because of warming, this metric will increase further.

Anyway things are looking pretty dicey.

And we have to reduce CO2 impacts pronto:

According to the latest available science, achieving the long-term temperature goal would require global greenhouse gas emissions to peak by 2020 and subsequently be reduced to zero before the end of the century. To limit warming to 1.5°C, this reduction to zero must take place around 2050

But Smil doesn't think this future is likely, because there's really no reason to imagine it would be so. We're still increasing the absolute amount of CO2 emissions. The relative ratio of fossil fuels is declining, but the absolute amount of energy generated is increasing. The population is still projected to keep growing to 9 B by 2050. And as countries in Asia and Africa become more prosperous (hopefully), the energy footprints will increase. Per capita, the middle class in America take 500 GJ per year, compared to per capita in the developing world at 20 GJ per year, an order of magnitude less.

(Kinda bored of the environment section, moving on)


Smil is highly critical of singularity believers like Ray Kurzweil. He’s also critical of “endless progress” narratives like those of Steven Pinker. And those that try to thread the needle between the two like Noah Yuval Harari (see Homo Deus by Yuval Noah Harari).

Instead Smil aligns himself in between techno optimists and systems thinkers with an understanding of natural limits like Meadows (see Thinking in Systems by Donella Meadows).

Rather than make predictions, Smil attempts to give a lower bound on the fastest possible time to make future transitions.

Projections of population are still like 9+ billion for 2050. These projections may be conservative but still likely to be wrong. Even though they are on S curves and not naive projections, S curves are hard to predict (see Forecasting s-curves is hard). And also, this does not take into account any discontinuous events and black swans.

Random things to look into:

  • Pumping water up-hill as giant battery: Apparently the efficiency of storage of a large water reservoir that is artificially pumped up is extremely good, an only 25% loss from this, and stores energy extremely well. Far better than any chemical battery could.
  • Look for energy density comparison (in watts per kg) of batteries versus kerosene for air travel and diesel for large naval engines.
  • Apparently the energy footprint of wood vs plastic is heavily in favor of wood. What are the implications of that?

Collective action problems: The global dysfunction in the back of global wisdom Homosapien sapiens is a major barrier to reducing climate impact. In general the collective action problem is basically unsolved as we can see in the 21st-century. Related to Collective action problems aka social dilemmas.

Smil ends with a #2x2:

|            | Simplicity     | Complexity          |
| ---------- | -------------- | ------------------- |
| Minimalism |                | Smil's prescription |
| Maximalism | High modernism |                     |


  • Favor a multitude of approaches rather than relying on a single perfect solution (eg. No magic geo engineering bullet)
  • No place for a priori ideological purity. No insistence on what is best. Pragmatism (eg. Open to nuclear)
  • A hyper-focus on electrification may be over doing it (ahem Rewiring America by Saul Griffith). Avoid killing hydrogen vehicles by overtly favoring EVs with excessive rebates and ownership targets.


  • Systemic small changes can help a lot. Not all fruits need to be available year round. Not all cars need to be SUVs.