Boris Smus

interaction engineering

Why Information Grows by Cesar Hidalgo

Why Information Grows starts off strong with an accessible introduction to information theory and some interesting physics-inspired ideas. Sadly, as the book proceeds, it becomes clear that the author has pieced together multiple essays that were and remain loosely connected. It's clearly a derivative work, often far less clear than the books that are cited.

That said, a few things stood out for me:

Information is the opposite of entropy

Information is stimuli that has meaning in some context for its receiver. There is some relationship between information and complexity and information and order but Hidalgo dances around this question.

Information in the universe is fractally distributed

Information is rare in our universe. As far as we know, Earth is one place where information is highly concentrated. Earth is to information like black holes are to matter. This observation struck me as very anthropocentric. Maybe I'm reading too much science fiction!

On Earth, information is not uniformly distributed. Specific countries, companies, and individuals seem to represent more of it.

Rubic's cube example from an info theory perspective

The author does a decent job of explaining basic information theory. I liked his Rubic's cube example:

  • There are e19 different states in the system.
  • The most information is in the lowest entropy state, the solved solid colored state
  • You can’t just jump to the solved state without cheating
  • It takes under twenty moves to solve a cube from any state
  • Most algorithms take longer than the theoretical optimum sequence of moves.
  • Most random moves cause entropy to increase.

Information state changes are constrained

The Rubic's cube can't go from any state to any other state unless you cheat and detach the cubes from the cube's core. The only valid moves are rotations. This is a profound point that I think is super relevant for any policy discussions. Often it’s very clear what a desired state of affairs is, but very unclear how to get to there.

Encoding information

Application to DNA: interesting sequences that code for something important are usually the least random ones. (Really?) Notably these information rich states don’t tell us about what this DNA means. Meaningless forms of order.

Information itself is meaningless. But can be absorbed my people to form knowledge. And practice can lead to know-how which I think is same idea as métis as in Seeing Like a State by James Scott and сноровка.

Energy flows cause spontaneous order

Entropy only increases in a closed universe. Meaning that this universe has no flow of energy into or out of it.

Ilya Romanovich Prigogine, a Russian Jewish born Belgian studied Dissapative systems, which are systems far from thermodynamic equilibrium, with substantial steady-state flows of energy. Earth is such a system because the sun is the source of all our power.

In situations where there is a steady state, there is a tendency for spontaneous order. A whirlpool will spontaneously appear when you open the drain to a bath tub. Are there other simple examples?

This is super cool! Imagine all of life as a sort of very complex whirlpool. Actually, Ted Chiang kinda did in a more poetic version of this idea found in the story called Exhalation. Exhalation by Ted Chiang

Information can be stored in solids

Information can be preserved, and this preservation is best done in a solid state of matter, especially if it is non-uniform (eg. not a simple solid crystal, more like an organic molecule). The canonical example of this is DNA.

I didn't really like Zen and Art of Motorcycle Maintenance, but this quote works well here:

That's all the motorcycle is, a system of concepts worked out in steel. There's no part in it, no shape in it, that is not out of someone's mind - Robert M. Pirsig

Goldilocks zone for information to appear and remain

Interesting analogy to physics: a desirable temperature range. you need liquidity for information to grow (eg. Whirlpool example, as well as the notion of liquid networks from Where Good Ideas Come From by Steven Johnson). But if it’s too liquid, solids can’t form to store that information.

There are no long chains of proteins in the sun

Energy is needed for information to emerge and solids are needed for information to endure. For information to explode we need the ability of matter to compute.

Simple computation, very broadly defined

Hidalgo defines computation extremely broadly. You don’t even need life for computation to occur in the natural world. He introduces a simple example of computation, a chemical clock, in which a set of regents can be mixed together to make a reaction that transitions in states at predictable time frames. Trees also compute in a similar way.

I don't really see what practical connection there is to computation the way we think of it more colloquially.

Products as crystallized imagination

Everyone uses toothpaste. Very few know how to produce sodium fluoride, or how to get it into that tiny tube. Toothpaste the product gives billions of people access to the practical benefits that a small group of toothpaste inventors created. Products allow us to transcend our individual limits. Toothpaste gives everyone the ability to maintain our teeth into old age.

Crystallized information comes in many forms. Most of human activity is like this. Writing, speaking, music making, dancing. But Hidalgo only cares about products. Products empower others with practical uses of your ideas.

The economy then is the substrate that allows products to spread from one individual to another.

I'm mildly annoyed at the author for:

  1. Not really justifying why he's so obsessed with products specifically. From an information growth perspective, does art not also compound?
  2. Calling these things "products". It's very capitalist-y for no good reason. I strongly prefer artifacts.

The bad stuff

The first part of the book discussing information theory and Prirogine's ideas was quite good.

Next, Hidalgo tries to apply information theory to people, companies, countries, and the global economy. This gets increasingly awkward.

Personbytes, a useless analogy

The personbyte is a transparently info theory derived quantity that's supposed to somehow represent a person's raw potential, I guess? The analogy is not great but illustrates a point. The information-theory notion of a byte is clean and mathematically precise. Applying it to people seems oddly reductive. People vary hugely in their potential and are products of their culture which is perpetually evolving. Stripped of this, we are just smart monkeys. My personal preference is to not introduce this faux-information theory concept. The same points could be made better without it.

Hidalgo's ultimate point is well taken: that the crystallized imagination encoded in artifacts increase your own imaginative potential. This puts you in a sort of a network with other people, and being part of this network lets you create more complex output. It’s much easier to build an electric scooter if your job is to put pieces together rather than inventing battery, electric motors, aluminum welds etc.


Hidalgo continues down this road and introduces the firmbyte, a unit of quantization for companies. Similar critiques apply.

Ford once had a plant called The Rouge that created cars end-to-end, taking iron ore, soy beans, rubber as input, then performing the whole manufacturing process (refining, smelting, forming, cauterizing, etc). This facility spanned an area roughly 2km x 2km and employed 100,000 people. This isn't common practice anymore.

Manufacturers don’t generally own the whole production line from raw material to final product. There is a natural unit size for a company, and that is the firmbyte. Again, extremely fuzzy, maybe inspired by information theory. Do not want.

The point here is even weaker: some things require multiple companies, or firmbytes to create. If there's too much friction to doing deals between companies, large networks can't be formed. Free trade agreements and standardization reduce friction between companies.

Imagination trade imbalances

Blending the idea of crystallized imagination and global trade, Hidalgo considers trade imbalances from the perspective of imagination. In the trade between Korea and Honduras, there is an imagination trade imbalance, with Korea providing in the commodity of imagination. He then ranks countries in terms of how much information capital they produce. Again, meh.

Hidalgo touches a bit on the Resource curse. This would be interesting to understand in more detail.

More derivative stuff

In the spirit of being a random amalgamation of cool ideas Hidalgo likes: