What this covers

César Hidalgo, a physicist-turned-complexity researcher, speaks with Sean Carroll about how information—understood as physical order sustained by energy flows—shapes the structure and inequality of economies and societies. The conversation moves across physics, economics, and history to argue that the capacity to compute or make things, rooted in Knowledge vs Information, is what drives complexity. Hidalgo draws on Ilya Prigogine's work on dissipative structures to ground information in thermodynamics, then traces how this framework illuminates why places become unequal, why innovation clusters geographically, and how communication technologies reshape what societies remember and value.

The episode covers considerable ground. Hidalgo distinguishes Tacit vs Explicit Knowledge—arguing that what cannot be codified or transmitted through instruction, like Michael Jordan's basketball skill, resists movement and therefore concentrates spatially, creating temporary monopolies and reinforcing regional inequality. He examines Geographic Knowledge Diffusion through cases like Ecuador's failed Yachay science city and successful hubs like Silicon Valley, showing why spending money alone cannot seed innovation where knowledge networks do not exist. He also tracks how Communication Technology Eras—from writing and printing through film, radio, and television—have reconfigured which kinds of people become globally famous, and explores Economic Complexity as a driver of both inequality and equality, depending on whether you compare across countries or within them, producing a subtle Simpson's Paradox. Along the way, he discusses the role of migration in diffusing talent, how vertical integration made sense for Henry Ford but no longer does, and what mentorship actually transfers.

Sharpest takeaway

Hidalgo argues that information physically embodied as order, combined with the human capacity to 'compute' or make things (knowledge and know-how), is the fundamental driver of complexity in economies and societies, and that because tacit knowledge is hard to move, it concentrates geographically and shapes inequality, innovation, and collective memory.

  • Information is physical order that grows where energy flows sustain out-of-equilibrium structures, as Prigogine showed
  • Tacit knowledge is socially embedded and hard to move, creating geographic concentration and temporary monopolies
  • Communication technologies (writing, printing, film, radio, TV) reconfigure how societies create and remember knowledge

The argument · threads9 threads · 34 claims
0.65

Complex economies generate diverse, competitive activities that enable middle-income jobs and reduce inequality.

4 pointscentrality 5/5
  • More complex and developed places (cities, advanced countries) have greater diversity of economic activities in a nested structure: places that do few things tend to do only things common to places that do many things, so undiversified economies are stuck with a limited and redundant, uncompetitive offering, which is a mechanism behind inequality since anyone whose capabilities are commonplace cannot command a decent income.

    “diversity grows with the level of development and complexity so that the places that do few things, they not only do few things, they do things that are common to the places that do many things.”

  • Division of knowledge is distinct from division of labor: a division of labor can have many people doing the same task (e.g., warehouses of seamstresses producing identical shirts, yielding economies of scale), whereas division of knowledge has people doing different things and passing inputs to each other to create products no individual could make, as in aircraft or pharmaceutical manufacturing where complexity is measured in 'person bytes' of knowledge.

    “the division of knowledge is quite different because we're doing different things, and we're passing on those inputs to each other as a way to create things that would be impossible for each one of us to do.”

  • Countries dependent on mineral or fossil-fuel resources cannot reduce inequality through redistributive policies and social safety nets alone; because hierarchical resource industries do not generate middle-income jobs, they must sophisticate their production structure to create the diverse, creativity-oriented activities that produce those jobs.

    “Because those middle income jobs are not going to happen just through the redistributed policies. You do need policies and social safety nets, but you do need to change the practice structure.”

  • Using measures of the knowledge intensity of economies, more complex economies tend to be more equal and less complex economies more unequal across countries (because resource-based industries like mining are hierarchical and regimented rather than creative), but the relationship inverts within countries (more complex cities like New York are more unequal), producing a Simpson's paradox driven by talented migrants concentrating in complex cities.

    “economies that are more complex tend to be less unequal... And economies that are less complex are more unequal.”

0.65

Knowledge is experiential, socially embedded, and geographically circumscribed, making it far harder to move than materials or capital.

4 pointscentrality 5/5
  • Knowledge is hard to move long distances because it is socially embedded, and social networks are geographically circumscribed; barriers such as language and cultural differences further limit the creation of links between people, which extensively limits knowledge diffusion and trade.

    “the reason why it's hard for knowledge to travel long distances is because it's socially embedded. It's actually that social networks are geographically circumscribed, and that limits the diffusion of knowledge.”

  • To produce things you must combine materials, technology, and knowledge, and because the world minimizes cost it moves whatever is easiest to move; knowledge is by far the hardest to move because it is experiential and embedded in large teams, not equivalent to encoded information, which is why production concentrates geographically and why it is cheaper to ship raw materials to where knowledge already exists (e.g., Chilean copper to Korea for batteries) than to recreate the knowledge.

    “the hardest factor to move is knowledge.”

  • Explicit knowledge is anything that can be communicated and codified through an act of communication, such as a recipe in a cookbook, while tacit knowledge cannot be effectively transmitted by communication and requires practice; the classic example is that a multi-day seminar by Michael Jordan would make you only epsilon better at basketball because that skill is tacit, and the world is full of tacit knowledge that is hardest to see precisely because it cannot be codified.

    “Explicit knowledge is all of the knowledge that I can communicate through an act of communication... The experience of having cooked with the chefs from El Bulli or any famous restaurant is something that is much harder to communicate, so it would be considered tacit knowledge.”

  • Information is what is encoded (an organized structure like DNA or a car), while knowledge is the computational capacity to make, transform, recombine, or reproduce that encoded information; like a recipe is useless without a kitchen and cook, encoded information is inert without the machinery and know-how to act on it, and this capacity to make is the true mystery and the thing hardest to accumulate at both biological and social levels.

    “information is what is encoded, and knowledge is this ability to make.”

0.51

Geographic concentration of production follows from knowledge's immobility; whoever discovers something first enjoys temporary monopoly until knowledge spreads.

4 pointscentrality 4/5
  • Because products diffuse quickly but the know-how to make them diffuses slowly, whoever first figures out something people want enjoys a temporary monopoly until the knowledge spreads, which explains geographic clustering such as carmaking concentrating in Detroit rather than spreading nationally, and Silicon Valley and China holding monopolies over the markets and technologies they pioneered.

    “If you figure out something that people want, you're going to have that monopoly because the product is going to diffuse very quickly. But the ability to make it is going to diffuse very slowly.”

  • Ford's River Rouge plant achieved extreme vertical integration—taking in soybeans and metal at one end and producing cars at the other—because transportation was poor at the time, so concentrating the full value chain in one location was necessary to achieve the economies of scale needed for affordable vehicles, whereas today distributed international value chains make such monolithic models obsolete.

    “at the time of Henry Ford where transportation was not that good... You needed maybe to produce that level of vertical integration to be able to get the economies of scale that he needed to produce affordable vehicles.”

  • You cannot create a knowledge hub anywhere simply by spending money, because the process of knowledge diffusion constrains where knowledge can be created; Ecuador's Yachay project spent roughly 1% of GDP (a billion dollars) building a science city on isolated agricultural land far from existing knowledge centers (Quito, Guayaquil), and it failed, churning through six or seven presidents and costing roughly a million dollars per student—enough to have put each student through Stanford for life.

    “Basically was this idea that if you had enough money you could create knowledge anywhere. And that's not true because their process of diffusion that constrained the creation of knowledge.”

  • Trade barriers against Mexico are misguided because modern automotive value chains are internationally integrated and disaggregated rather than vertically integrated, as evidenced by Mexico being the number one exporter of beer due to manufacturing for US brands.

    “that's for instance why [Trumptards] are stupid when it involves Mexico because the US value chain and automotive sector”

0.46

Complex systems require nested collective structures to transcend individual capacity limits and accumulate increasing orders of complexity.

2 pointscentrality 4/5
  • Complexity and order are not everywhere in the universe but concentrate in places like Earth because out-of-equilibrium systems sustained by energy flows self-organize into dissipative structures, following Prigogine; life requires continuous energy consumption precisely to stay out of equilibrium and maintain its level of organization against entropy.

    “those systems that were out of equilibrium tended to self-organize into what he called the dissipative structures.”

  • Every system has a finite capacity to accumulate knowledge and know-how, so the only way to transcend that limit is to form nested collective structures: single cells form multicellular organisms, humans form teams, teams form organizations, organizations belong to industries and cities, each level renormalizing the units that hit a ceiling to accumulate more complexity than any unit could alone.

    “Humans transcend that capacity by forming teams, teams transcend those capacities by forming organizations, organizations transcend those capacities by belonging to industries and also by having cities”

0.46

Cities concentrate wealth and innovation yet create long commutes and housing scarcity, generating unsustainable social pressure without dense infrastructure.

2 pointscentrality 4/5
  • Since complex economic activities concentrate in space even more than simple activities and wealth is increasingly generated in cities, the US is failing because its major cities cannot accommodate more people due to infrastructure and building constraints; without dense, livable neighborhoods and good transportation, exclusion and three-hour commutes will generate unsustainable social pressure, as seen when social unrest is triggered by rising gas or transport prices.

    “the major cities of the US are failing to accommodate more people because of problems with infrastructure, problems in the ability to build and so forth.”

  • Average travel time increases with city size despite greater density and nearby destinations, so big cities like New York, DC, and LA have long average commutes, while commuting alone (driving) is characteristic of more rural areas, a pattern visible in Data USA's commute maps.

    “when you create the average travel time map, boom, all of the cities light up. When you create the commuting alone map, then it's the opposite.”

0.44

Knowledge hubs cannot be created by spending alone; they require proximity to existing knowledge centers and social networks.

1 pointcentrality 4/5
  • The most valuable thing to gain from a mentor is not technical detail (which can be figured out alone) but tacit skills like judging when a problem is relevant, how to communicate to different audiences, and how to manage a career; these are best learned as a humble apprentice of someone who is where you want to be, ideally in one's 20s.

    “what I wanted to get out of him is understand when a problem is relevant. How should I communicate to people?”

0.44

Information and complexity require stability balanced with flexibility; order grows naturally where energy flows sustain dissipative structures.

12 pointscentrality 2/5
  • TikTok may be more social and family-oriented than prior peer-to-peer social media because it is watched together (as a collective couch experience requiring negotiation and discussion) like older television viewing, contrasting with Twitter and Facebook which are used alone and are isolating.

    “When I watch TikTok, I'm in the couch with my wife and my daughter and the three of us are looking at the same screen at the same time and talking about what we're watching.”

  • A large crowdsourcing study (Place Pulse: ~4,000 images rated by over 100,000 people) found that people's perceptions of whether a place looks safe, lively, or beautiful are highly transitive and near-universal, because differences between images are so large they overwhelm differences between people, allowing perception to be measured at scale—and these perceptions can be independent of whether a place is actually safe.

    “people's preferences were very transitive... The difference between images is so large that it overwhelms the differences between people.”

  • TikTok is distinctive as the first global social medium popular in both China and the West, and unlike prior peer-to-peer social media it is fundamentally broadcasting (a passive scrolling 'for you' feed rather than communication or effective search), effectively reinventing television for the phone era with high-quality 15-30 second content.

    “It's the first global social media, because it's a Chinese company. It's the only social media that is popular in both China and the West... It's not based on peer communication. It's actually broadcasting.”

  • Hidalgo is optimistic because, just as information and order grow on Earth since the ability to create order and complexity exceeds the rate at which it is destroyed, positive things in life tend to add up and build on each other better than negative things do.

    “our ability to create order and complexity is larger than the rate at which it's getting destroyed. I think that's also true for many of the positive things.”

  • Combining computer-vision perceptual maps with mobile-phone movement data and controlling for distance to subway, business district, and density of jobs and population, researchers found people tend to avoid unsafe-looking places, an effect stronger for women and the elderly but reversed for people under 30—especially young men—who tend to hang out in unsafe-looking places.

    “we find that like people tend to avoid unsafe looking places... Those effects are modulated by demographics. It's stronger for women and for the elderly, but it's reversed for people below 30.”

  • Preserving complex information requires stable carriers like DNA crystals, but perfect stability would prevent growth; complexity requires the ability to explore configuration spaces, so DNA must be stable enough to preserve information yet flexible enough to change, because complexity requires diversity.

    “You need to be able to explore those spaces of configurations so that you can actually then grow in complexity because complexity requires diversity.”

  • The hard but valuable task in data tools is not providing files but integrating 15-30 disparate datasets into narratives, transforming data into stories with semi-algorithmically generated text and visualizations, so that even a small US census place has a complete profile of 70+ visualizations, making the data easy to use even though it is hard to build.

    “What we do is we transform data into stories. That's the main form of integration that we provide.”

  • The lasting value of a physics education is the habit of always maintaining an interplay between theories and experience to understand the world, a duality useful in physics, economics, and most other fields.

    “to understand the world, you need to always have an interplay between theories and experience.”

  • Complexity in economies, societies, and ecosystems does not grow monotonically but can collapse rapidly, as in ecosystem collapse under climate change or sudden social unrest in countries thought to be doing fine, such as Chile during the week of recording.

    “There are countries that sometimes everybody thinks that they're going fine, and things turn very quickly, like what is happening in Chile this week”

  • Beyond matter and energy/momentum, there is a third basic quantity needed to understand the universe: information, defined as how things are arranged or ordered rather than what they are or how they move.

    “there's a third quantity that we need to consider, which is not things nor how they're moving, but how they're arranged or ordered. To me, that's the basic idea of information.”

  • Musicians could earn large incomes only during a narrow historical window because recording technology (records, tapes) trapped music in physical media, allowing it to be sold at high margins; before recording, musicians could only perform live and could not monetize widely, and today, when music can no longer be trapped in physical media, monetizing it is again nearly impossible.

    “those records allowed to constrain the diffusion of music in such a way that you could monetize it because music was trapped on those discs.”

  • From a pure Shannon perspective, a random sequence contains the most information because it is the hardest to predict, even though it means nothing; in a communication context, by contrast, the most informative words are those that most reduce the listener's uncertainty about what is being said.

    “If we're thinking from a pure Shannon perspective, basically a random sequence is going to be the one that contains more information because... it's the hardest to predict.”

0.38

Communication technologies reshape the composition of collective memory and what kinds of people gain global fame.

4 pointscentrality 3/5
  • New communication technologies change not just the size but the composition of collective memory: before printing, globally famous people were mostly political and religious leaders; printing introduced famous artists and scientists (painters, composers, astronomers, writers); film and radio shifted arts fame from composers to performers (musicians, singers, actors); and television created the fame of sportsmen, who require live, time-bound viewing.

    “when new communication technologies were introduced, the composition of our collective memory changed, not only the size.”

  • Human history can be understood as a sequence of communication-technology revolutions that reconfigure how we create and process information: the cognitive revolution (language, ~80,000-100,000 years ago, including the ability to discuss fiction), the agricultural revolution and cities (~10,000 years ago) enabling cross-generational knowledge accumulation, the writing revolution (~700 BC, peaking in ancient Greece), the printing press, and later film, radio, television, and the internet.

    “I like to think of the history of our planet or of our last thousands of years as a history of changes in communication. Technologies that have reconfigured the way that we create and process information”

  • The printing press was the first urban, scalable, IP-intensive economic activity: before printing, wealth required land, serfs, mineral resources, or looting, but printing let a small urban team produce content cheaply at scale and become rich quickly, evidenced by the number of printers per capita in Europe stabilizing within only 50 years, making it effectively the first form of mass production.

    “printing is the first economic activity that is urban and scalable.”

  • Mapping a global language network (nodes are languages linked when likely to be co-spoken or translated, built from book translations, Twitter, and Wikipedia edits), the centrality of a language in that network predicts the number of globally famous people it produces better than the language's population or wealth, because the language network is the most aggregate version of the global social network and fame depends on social connectivity.

    “the centrality of a language in that network explained the number of famous people produced by the language better than the population of the language, better than the wealth of that language.”

0.30

Migrant talent is disproportionately concentrated among the most educated and inventive, driving innovation and job creation globally.

1 pointcentrality 3/5
  • Migration is a major channel of knowledge diffusion and is steeply biased toward the most talented: roughly 1% of people without a college degree migrate, about 5% of those with degrees, about 10% of inventors who have filed a patent, about 31% of Nobel Prize winners, and since the 1970s about 60-something percent of US Nobel laureates were immigrants, so a globally mobile tail of talent drives innovation and job creation, and the US has historically captured about 50% of all migrating PhDs.

    “people without college degree, about 1% of them migrate. People with a college degree it's about 5%. Inventors that have filed a patent about 10%. Nobel Prize winners about 31%.”