Key Takeaways
Measure innovation by adoption, not by the cleverness of the invention
Ideas need systems. Smil distinguishes invention, the creation of a tool, material, machine, or production method, from innovation, the process of introducing and mastering it. Michael Owens’s bottle-making machine, for example, raised output from about 200 bottles an hour in semiautomatic production to 2,500 without workers handling hot glass. Yet a clever device matters only when organizations can put it to work.
Diffusion changes economies. The Soviet Union produced important scientific and military inventions but lagged in modernizing steelmaking and supplying ordinary consumer goods. China’s rapid growth, by contrast, relied heavily on adopting and scaling technologies developed elsewhere. Its first commercial transaction after the 1972 opening to the United States was a purchase of advanced ammonia plants, an industrial technology crucial to fertilizer production.
What's striking is how this distinction shifts attention from the lone inventor to the surrounding capacity: trained workers, factories, finance, institutions, and access to know-how. A patent records a claim to novelty, not proof that a society can build, maintain, or afford the thing. The contrast also warns against treating invention counts as national scorecards. One qualification matters: the boundary between invention and innovation is porous, since adapting a foreign technology can itself require substantial invention. The useful question is therefore not who had the first idea, but who made it reliable, affordable, and widely useful.
A useful invention can become harmful when its side effects arrive late
Benefits can conceal costs. Leaded gasoline improved engine performance, DDT killed disease-carrying insects, and chlorofluorocarbons, or CFCs, made household refrigeration safer than earlier toxic or flammable refrigerants. Each addressed a real problem. Each also created harm that was difficult to see at first, though not always impossible to foresee. Lead’s toxicity was known early; the ozone damage caused by CFCs became clear only after decades of atmospheric research.
The stories differ. DDT rapidly reduced malaria and typhus, but broad agricultural spraying spread a persistent pollutant and selected for resistant insects. Lead emissions harmed children over generations. CFCs accumulated far above the ground, where their effects on ozone were discovered long after their commercial success. These cases are not interchangeable: they differ in what was known, who benefited, and how the damage emerged.
The cases make a strong argument for tracking effects beyond the place and time where a product is used. A refrigerant can escape into the atmosphere; an insecticide can move through food webs; exhaust can expose children far from the factory that made the additive. This is a problem of delayed feedback: early commercial success is weak evidence of long-term safety. The distinction among cases is equally important. Treating every unintended harm as unforeseeable excuses poor decisions, while treating every later discovery as obvious in advance erases genuine scientific uncertainty.
Do not let a cheap technical fix silence better alternatives
Convenience won the contest. Engine knocking can damage a motor, so General Motors searched for an additive that would let engines run at higher compression and use fuel more efficiently. In 1921, Thomas Midgley’s team found that a small amount of tetraethyl lead worked. Ethanol and benzene blends were also known options, but they posed different production and supply challenges. Lead was cheap, effective in tiny quantities, and could be controlled through a corporate patent.
The warning was not hidden. Physicians raised concerns, and lead poisoning among workers at a processing plant included five deaths in 1924. Industry promoted the additive under a name that obscured its lead content. Later, millions of tons of lead entered the environment through US gasoline use. The phaseout began largely because lead damaged catalytic converters needed to reduce smog, not because health concerns finally prevailed.
This episode shows how a technology can become locked in through convenience, ownership, and infrastructure rather than through clear superiority. The additive solved a measurable engineering problem, while its public-health cost was diffuse and fell heavily on children who had no role in choosing it. That mismatch helps explain why warnings struggled to compete with claims of economic necessity. The alternative question is crucial: when a proposed fix is selected, what other routes are being abandoned, and who has the authority to compare their full costs?
Environmental treaties work when evidence, substitutes, and industry align
A global problem required coordination. CFCs were stable, affordable refrigerants that enabled refrigerators and air conditioners to spread. In the 1970s, researchers showed that chlorine released from CFCs could destroy stratospheric ozone, which shields life from damaging ultraviolet radiation. Because these compounds persisted and crossed national boundaries, individual countries could not solve the problem alone.
Replacement made action practical. The 1987 Montreal Protocol began a global phaseout, later strengthened by amendments. Producers helped supply substitutes, including hydrochlorofluorocarbons and hydrofluorocarbons. Those replacements reduced or avoided ozone damage, but some contributed to global warming. The transition therefore did not discover a perfect refrigerant. It reduced one major danger while creating pressure to improve the next generation of cooling technologies.
The Montreal story offers a useful model of environmental governance: scientific evidence identified a mechanism, monitoring made the change observable, and available substitutes gave governments a path beyond prohibition. Industry participation helped make the transition feasible, though it does not mean commercial interests and public interests were identical. The remaining climate cost also exposes a common policy trap: solving a problem by moving it elsewhere. Strong regulation needs follow-through, including attention to replacement products, discarded equipment, and the energy used by the technology itself.
Judge transport revolutions by economics, not by speed alone
A spectacular machine can lose to a better system. Airships once seemed ideal for long-distance travel. The Graf Zeppelin carried more than 13,000 passengers and completed 144 intercontinental trips. The Hindenburg disaster became an enduring image, but airships were already losing their advantage as airplanes grew faster and more capable. The shift was not simply a verdict on one crash; the airplane’s performance was changing the competition.
Supersonic flight faced different limits. Concorde cut travel time across the Atlantic, but only twenty aircraft were built, and just fourteen entered commercial service. Its narrow cabin, limited range, high fuel use, operating costs, and disruptive sonic boom constrained demand. Governments subsidized the service, and the aircraft never displaced ordinary jetliners. Speed was valuable, but not valuable enough to offset the whole package of costs and constraints.
These histories resist a simplistic story in which each new mode naturally replaces the old one by going faster. Transport is a network service, so the vehicle must fit routes, airports, infrastructure, passenger demand, and operating economics. A technical advantage matters only when it improves the journey people actually buy. The comparison also shows why failure should be diagnosed carefully: the Hindenburg tragedy is memorable, but a dramatic accident can obscure the slower commercial forces already making a technology less competitive.
Nuclear power became a useful failure, not the promised electricity dominator
The promise outran the market. Nuclear fission was promoted as a source of abundant, cheap electricity, and some forecasts anticipated thousands of reactors. By 2020 it supplied about 10 percent of global electricity, with more than 400 reactors operating. In the United States, the first commercial program adapted a naval submarine reactor design, a politically expedient choice that helped establish the industry but also narrowed its options.
Costs and expectations diverged. Electricity demand slowed, construction took longer, and expenses rose. Utilities canceled projects, and proposed breeder reactors, which were meant to produce more usable nuclear fuel than they consumed, failed to become practical commercial systems. Yet the record is not simply failure: reactors provide dependable, low-carbon electricity in several countries. Smil calls this a successful failure: a real and valuable technology whose achievements fall far short of its extraordinary promises.
The phrase successful failure is useful because it prevents two opposing distortions: dismissing nuclear power as worthless, or treating its existence as proof that its original promises were fulfilled. The case also shows how early design choices can create path dependence: once investment, expertise, and regulation gather around one system, alternatives become harder to pursue. A fair assessment must compare actual performance with both stated ambitions and available alternatives, while recognizing that a technology's value can differ sharply across countries.
Two centuries of prototypes make hyperloop promises less, not more, credible
Old ideas can keep returning. Proposals for passenger travel through low-pressure tubes date to the early 1800s. Later inventors added magnetic levitation and proposed vast networks, yet no complete long-distance system has entered service. The modern Hyperloop concept inherited the same basic challenge: building and maintaining a long, safe tube, keeping pressure low, and moving passengers reliably at high speed. A Nevada test pod reached 175 kilometers per hour, far below promotional visions of near-sonic travel.
Other quests face their own gates. Nitrogen-fixing cereal crops could reduce fertilizer pollution, but making staple grains work like legumes remains unresolved. Controlled nuclear fusion has achieved important laboratory results, yet its experimental devices do not generate saleable electricity. In each case, a scientific or engineering milestone is only one step toward a system that can be operated, financed, and widely adopted.
These examples suggest a practical rule for evaluating a promised breakthrough: ask what stage has actually been demonstrated. A mechanism that works in a laboratory is not yet a durable process; a prototype is not a safe transport network; and an experimental reactor is not a power station. The obstacles vary, from plant biology to plasma physics to civil engineering, so no single timeline applies. Repeated promises deserve skepticism, but skepticism should target unsupported certainty, not the value of continued research.
Microchips accelerated; the systems feeding civilization mostly did not
One spectacular curve misleads. Microprocessors improved at extraordinary rates after their introduction. By 2020, chips held more than ten million times as many components as the first microprocessor. But this pace is not typical of food, energy, materials, or transport. Crop yields often rise by only 1 to 2 percent a year, commercial batteries improved about twelvefold in energy density between 1900 and 2022, and jetliner cruising speeds have barely changed since the late 1950s.
The foundations came earlier. Smil points to the 1880s as a remarkably productive decade: electric power systems, transformers, steam turbines, electric motors, and early automobiles took shape. Modern computing depends on electricity, steel, agriculture, and transport systems that were already developing before microchips. Electronics improved how those systems are managed, but it did not create all their material foundations.
The comparison corrects a perceptual bias: digital change is visible, fast, and easy to count, while improvements in infrastructure and production are slower and less dramatic. That does not make computing unimportant. It means that extrapolating its growth curve into every other field is a category error. Physical systems face constraints of materials, land, construction time, and coordination that information processing can sometimes help manage but cannot simply erase.
Spend first on spreading proven essentials, then on speculative salvation
Deployment can matter more than novelty. Many people still lack dependable access to food, clean water, electricity, education, and health care. Smil argues that much progress in these areas depends less on spectacular invention than on extending methods already known to work. The same principle applies to antibiotics: careful prescribing and avoiding routine mass use in livestock can help limit resistance without waiting for a miracle drug.
Hard problems take time. Cancer mortality in the United States fell 27 percent from 1999 to 2019 after decades of uneven research and treatment advances. Decarbonization is an even larger systems challenge. Fossil fuels supplied 83 percent of global primary energy in 2020, down only modestly from 87 percent in 2000. Reaching zero by 2050 would require a pace of change far beyond that recent record, alongside new infrastructure and technologies.
The argument is not that invention should stop, but that the prestige attached to novelty can divert attention from implementation. It is often easier to announce a moonshot than to fund maintenance, distribution, or public services. At the same time, realistic timelines should not become an excuse for delay. The useful distinction is between goals and forecasts: ambitious goals can guide effort, while credible forecasts must account for construction, supply chains, unequal starting points, and the time needed for systems to change.
Analysis
Smil's central contribution is to replace the heroic, linear story of invention with a systems history. The decisive event is rarely a clever idea alone. It is the joining of materials, manufacturing, capital, infrastructure, regulation, and public acceptance. That is why the book's examples span the whole life of a technology: a concept may be brilliant, commercially successful, socially harmful, or simply unable to scale. Those outcomes are not synonyms, and treating them as such obscures what history can teach.
The book is especially persuasive when it distinguishes different kinds of uncertainty. Leaded gasoline was introduced despite substantial warnings about toxicity, making its harm partly a failure of judgment and governance. CFC damage to ozone was harder to foresee, but became actionable through atmospheric research and international cooperation. Airships and supersonic transport were not primarily undone by a single accident; competing systems and costs weakened their case. Fusion, by contrast, remains an unresolved technical and economic challenge. These distinctions make the caution more useful than blanket distrust of new technology.
A further insight is that failure often begins as a measurement problem. Early success is easy to count: bottles per hour, aircraft speed, patents, or laboratory output. Harder to measure are delayed health effects, the cost of supporting infrastructure, or whether a prototype can operate economically for decades. The book repeatedly asks readers to widen the accounting boundary and lengthen the time horizon.
There is a limit to the method. A selection of cautionary cases cannot establish that innovation is generally slowing, nor can past failures determine which present projects will succeed. The strongest application is diagnostic rather than predictive: identify what has been demonstrated, who bears the risk, what alternatives were displaced, and what must still be built. That discipline preserves optimism while stripping it of automatic certainty.
Review Summary
Invention and Innovation by Vaclav Smil examines technological failures and successes, challenging overhyped claims about innovation. Smil explores inventions that became undesirable (like leaded gasoline), failed to dominate (like airships), and those still awaited (like nuclear fusion). He argues against unrealistic techno-optimism, emphasizing the slow pace of progress in crucial areas. While some readers found the writing dense or pessimistic, many appreciated Smil's data-driven approach and thought-provoking insights. The book offers a sobering perspective on innovation's challenges and limitations, encouraging a more realistic view of technological progress.
People Also Read
Glossary
Innovation
adoption and masterySmil uses innovation to mean the process of introducing, adopting, and mastering new materials, products, processes, and ideas. It is distinct from invention, which may produce a novel design or discovery without its becoming widely used. The distinction explains why a country or company can invent successfully yet fail to convert those ideas into productive capacity.
Successful failure
useful, but underdeliveredSmil's description of nuclear fission's mixed record. Fission became a substantial, dependable source of electricity in some countries, but it did not achieve the dominant, exceptionally cheap energy system its advocates once forecast. The phrase evaluates an invention against both its real-world contribution and the much larger expectations attached to it.
Haber-Bosch synthesis
industrial ammonia productionThe industrial process developed from Fritz Haber's laboratory work and Carl Bosch's engineering advances. It combines nitrogen from air with hydrogen under high pressure and high temperature, using a catalyst, to produce ammonia. Ammonia supports synthetic nitrogen fertilizers, which greatly increased crop yields but also contribute to pollution when much of the applied nitrogen escapes into the environment.
Diazotrophs
nitrogen-fixing microbesMicroorganisms able to convert atmospheric nitrogen, which most plants cannot use directly, into reactive nitrogen compounds. Some live freely in soil or water, while others associate with plant roots or live inside plant tissues. Research seeks ways to use these bacteria to supply more of the nitrogen needed by cereal crops and reduce dependence on synthetic fertilizer.
Tokamak
magnetic plasma confinementA doughnut-shaped experimental device that uses strong magnetic fields to confine extremely hot plasma for nuclear fusion research. The design originated in the Soviet Union and became the leading approach to magnetic fusion experiments. A tokamak can demonstrate important plasma behavior, but an experimental device is not automatically a practical electricity-generating power plant.
Q
fusion energy gain ratioIn fusion research, Q compares the thermal energy produced by fusion with the energy injected to heat the plasma. A value of 1 is called breakeven on that limited accounting. It does not include all electricity consumed by a facility, nor does it show that the system can generate continuous, affordable electricity. It is one performance measure, not a full commercial test.
Moore's law
chip density growth trendA description of the rapid historical increase in the number of components that could be placed on an integrated circuit, often expressed as a doubling roughly every two years. Smil uses it to illustrate how exceptional the pace of digital hardware improvement has been compared with slower advances in agriculture, energy, materials, and transport. The trend is not a universal law of technological progress.
Download PDF
Download EPUB
.epub digital book format is ideal for reading ebooks on phones, tablets, and e-readers.