Housekeeping note: This one gets a little dense, so I would recommend saving the footnotes for the end.
I. The System
Civilization has managed, through capitalism and democracy, to incentivize growth and redistribute value and wealth *adequately enough* in the past ~200 years.
Capitalism and democracy – or more fashionably, liberalism – has become the dominant system through which humanity has pulled off its amazing growth story. So for the rest of this essay, I’ll shorten this to simply, the System. This is not purely out of laziness – it’s literally the most important of all human-made systems, so it deserves the title. Let it serve as a reminder to its overarching import in all matters of human life, save for indigenous peoples removed from modern technology, who will a) technically experience the gravity of the System through their inability to migrate or expand, and b) never read this.
If there is a single key feature of the System, it’s that it’s mostly immovable: you can’t really undo stuff or introduce a top-down, paradigm shifting mandate – you can only change things at the margins.
II. Time
Each time there has been a major technological leap, the aggregate gains to society – measured through GDP per capita, infant mortality, literacy, lifespan, etc. – were so obvious that we managed through the occasional union strike or populist leader. Yes, we still have inequality of wealth and opportunity, but broadly speaking, technology = good.
First, some table setting.
The default way to measure something like “technological leap” is through total factor productivity, which is a numerical value used in economic models. This gets overly academic for a practical, qualitative discussion, though, so I prefer to think of productivity improvement as the decrease in person-hours to complete a given task. If you can get across town in 20 minutes on a bike or 2 hours walking, well, the bike represents a productivity gain. Same thing with fire and Excel.
The most misleading piece of the “technological leap” framing is the word “leap”. I’d argue that, when viewed from the human’s perspective, most of the oft-cited leaps were actually gradual implementations of a new capability. Most technologies – and all paradigm-shifting ones – took decades for the “S” shape of the technology adoption cycle to take form.
Decades – or even a single decade – gives us humans enough time to adjust and mold the way we live our lives. Worth noting is that these adjustments happen at the edges, the margins of the System. That’s important because it signals a promise to our future selves that things will still kinda sorta be the same…and so it’s safe to take a 30 year loan on a house or start saving for your kids’ college tuition, for instance.
III. Acceleration
It’s well documented our species’ failure to grasp exponentials and compounding growth. Or in plainer terms, acceleration. Today, most educated adults can at least identify when exponentials creep into a problem. I attribute this collective software upgrade to some combination of COVID, YouTube, Warren Buffet and Khaneman-Tversky.
But we still suck at calculating exponentials and abstracting the insights from the maths. If you ask people to apply compounding heuristics in everyday life, they’ll panic like a middle schooler who is prompted to suddenly USE one of the words from a vocabulary exam in a sentence, rather than regurgitating a definition.1
I say all this to prepare for an obvious statement: acceleration happens at the System level. That’s right – not just your index funds or that mythical farmer who tricked the king into paying him grains of rice as a function of squares on a chess board (2^64 is very large).
IV. Paradox, bits, atoms
Yet there’s a paradox: We are compounding, accelerating, but our planes and cars go the same speed as they did 60 years ago. Our buildings and art don’t seem that impressive. We haven’t sent a man to Mars.2
The System is multivariate and far too complex to give one satisfying explanation of why progress has stalled (and for what it’s worth, I seriously could not recommend going down this rabbit hole enough). The best single explanation is that the low-hanging fruit of life-changing innovations have been picked. It gets harder and harder to alter the physical world when our theoretical physics is stalled out, for example.
The acceleration is still happening, but it’s been constrained to bits, not atoms, thanks to Moore’s Law, the internet and cloud computing. And unlike the acceleration of the 20th century, which produced big, fast, physical things, the acceleration in computing mostly produced one thing: digital information. To quote Peter Thiel, we got 140 characters, not the flying car.
Information is defined by Claude Shannon, the OG of Information Theory, as “the minimum amount of communication required to uniquely specify a message.” If that’s too abstract, think of it this way: information is the essential data needed to understand something. It’s the foundation upon which more complex ideas and systems are built. Information can be a Chuck Norris joke tweet (which informs us of alpha-maledom) or it can be a flying car prototype (which informs us of design constraints in a future Jetsonian mobility infrastructure).
Digital information is basically what it sounds like – it’s centered in idea exchange; it’s visual and audible more so than tactile; it’s software. It’s why your home office in 2023 truly does feel like the Jetsons compared to your grandfather’s, but your car doesn’t.
My view is that the rates of acceleration in the physical and digital do not necessarily correlate. In other words, the big moments of progress in bits-space and atoms-space tend not to overlap; in fact, they are dependent on one another. Rather than move directly with one another, we should expect them to resemble a sort of step-function shape.
This step-function shape and dependency thesis has some historical backing, which I’ve illustrated in the chart below:
So far, all the big stuff falls squarely into the left or right side. And this makes sense when trying to resolve our paradox: information feeds physical growth, which makes us richer and feeds information growth, which feeds more physical growth. The most virtuous of virtuous cycles.
Naturally, the System incentivizes the greatest economic returns, which are a product of return (benefit) and investment (cost). So the magic happens, and we accelerate along the greasiest tracks and pick the next generation of low-hanging fruit.
So what’s the next generation?
V. Allen
The quick answer is “no one knows, and anyone who seems highly confident is either dishonest, delusional or both”. Maybe we stumble upon the Big Sticky Note and life changes forever.3
But if I were a betting man…
Hypothesis: AI – specifically knowledge generation AI, such as LLMs – will catalyze our entry into a new physical acceleration era.
To recap some of our big informational technologies:
Books made possible the storage and distribution of information
The printing press made possible the mass distribution of information
The internet decentralized the production of information and brought distribution marginal costs down to zero
Generative AI brought the marginal cost of information production down to (nearly) zero
What might zero-cost knowledge production bring to the physical world?
VI. Physicality
Physical acceleration requires more resources and more information – ie, better leverage over resources. This leverage comes in the form of engineering, which is downstream of fundamental research in physics, chemistry, biology, computer science, etc.
Making progress in fundamental science is exceedingly difficult. As an extreme example, there’s no interstellar travel fruit to pick while we are stuck reconciling quantum physics and general relativity. Those two theories themselves were once upon a time very difficult fruits to pick, and unfortunately, one of the bugs (fruit fly?) in our System is that we don’t sufficiently incentivize long-term, low-probability fundamental research.
The most honest thing I can do at this moment is leave the following placeholder:
But since that is a shitty reward for slogging through this essay, here’s a thought: Could we have a lot of incremental progress just by having robots do more stuff for us? After all, it’s purely an engineering problem (with perhaps a sprinkle of philosophy / sci-fi ethics).
I think the answer is yes, in which case we will have the usual decade(s) to ease into a new instance of the System, Robots Edition. We need time to build robots that could drive trucks safely (where are they, by the way?), make and deliver us pizzas, clean our houses better than the decrepit little disc-shaped turd that I sometimes see whizzing around in my living room, do our travel expenses and walk our dogs (or do people like walking their dogs? Always been confused by this).
The scenario in which generative AI unlocks the proliferation of robots seems plausible. I’ve admittedly skipped a few steps between “this new AI can write awesome stories and code” to “this AI engineered a Boston Dynamics level robot on its own”.
But again: The System is multivariate and far too complex to understand, and human advancement is incremental. The physical-non-physical feedback loop actually looks nothing like the clean illustration I laid out. If we were to make a more realistic diagram showing the progress between the digital and the physical “robots” it would have millions of tiny feedback loops, much like the rest of the human technology story. It only looks clean and simple when you zoom out sufficiently far enough to not see the little fluctuations.
And I’m not suggesting this is a natural next step, it’s simply one that makes sense to me and is fun to think about. I cannot stress enough: nobody actually knows.
VII. February 29
What happens, though, when the technological leap to the other side of the diagram happens overnight, or in a year? When the leap happens over 1-3% of a person’s lifetime, rather than the typical 10-30%? Seriously, what the fuck happens?
Again, we’re used to decades for this sort of magnitude change; the big gears shift and the big wheels turn and things tend to work out that way, because the System is good at absorbing technology, when looking at the absolute effects on the majority of people.
The system is immovable as far as we can tell. But is an overnight leap the proverbial six-sigma event that breaks the model?
[thought box]
A friend recently asked me, in the context of historical populist movements, if I thought a legitimate coup – French Revolution style – is possible today in a developed, G7ish country.
My immediate answer was “no way”. Obviously the powers-that-be have the physical and informational firepower to squash anything that remotely resembles a threat. Upon reflection, the correct is “yes!” for the same reason I had to include words like “mostly” in my assessment of the System’s immovability. It may be well below 1% probability, but it’s possible because we are clueless as to the second-order and third-order effects of an overnight leap.
What happens if oil becomes worthless tomorrow? Or better yet, how would the French aristocrat have answered the same question in 1500?
I make this point more illustratively than anything. I don’t know if it’s possible for a paradigm shift at the System level to occur overnight. I suppose the closest things we’ve seen in history are the world wars, but those feel more like “systemic shocks” than actual leaps or changes in the way people live with the physical world.
Frankly, it’s all terrifying and terrific. Terrifying because it may lead to massive upheaval and war. Terrific because it might help us escape this challenging level in the game of humanity that we seem to be stuck on. Like that video game level where it takes forever just to find the boss you’re supposed to beat.
Technology is the answer to beating this level. It’s the answer to beat every level.
A leap is coming, and the overnight leap will too, eventually. I don’t know if we’ll be alive for it. What a time.
I was so bad at vocabulary tests that I would usually skip any associated anxieties and accept the B or C on the exam itself, and have to deal with the uphill battle of getting an A in English by writing essays that meticulously followed the classic 1-3-1 rubric, each paragraph with one of those buzzy “transitions” like “Initially” or “Finally” or “In conclusion”, really more of a pseudo math puzzle than an actual essay, which drove me nuts even as a kid and resulted in me dropping easter eggs in the essay that the teacher would not catch (I don’t think) for my own amusement, and well anyways this probably explains why I write footnotes like this from time to time and why I was honestly a little disappointed with the first-gen ChatGPT when it came to explaining qualitative concepts – always in that godforsaken 1-3-1 format – but it’s cool now because I pay $20 a month for the heavy duty GPT-4 version and will never be in school again.
For some take home reading and listening:
Tyler Cowen / The Great Stagnation
More recent ones (neither of which I endorse): The Techno-Optimist Manifesto vs The techno-optimist’s fallacy
I’m too lazy to fact check that 3M accidentally came up with the Sticky Note when their glue machine lost power or something, but I like its romance.
https://twitter.com/rowancheung/status/1735314466147827910?s=21&t=iBVNikdu7cYSJO4ET6TUTQ