Institutions, Individuals and Intelligences

TL;DR

For most of the last century, we have traded freedom for friction. We accepted slow, rule-bound institutions because they bought us stability. In this century, that bargain has frayed. Complexity has risen, attention has been shattered, and individuals—with simple stories and direct channels— have outcompeted committees. We now have a third actor on the stage: intelligences. They don’t vote or hold office (well, axshually…), but they already set defaults, enforce norms in code, and decide what we see. The question isn’t just who governs?—it’s which layer governs which things, and with whose incentives?¹

Introduction

I’ve been living with a version of this post inside my head for a while now. But a recent post by Dean Ball helped crystallize my thoughts a little more clearly. Enough, at any rate, to get started with the first draft of this post. We shall see where we have reached by the end!

Here’s the excerpt from Dean’s post that helped me (although, as always, you should read the whole thing):

The law enforcement of the internet will not be the government, because the government has no real sovereignty over the internet. The holder of sovereignty over the internet is the business enterprise, today companies like Apple, Google, Cloudflare, and increasingly, OpenAI and Anthropic. Other private entities will claim sovereignty of their own. The government will continue to pretend to have it, and the companies who actually have it will mostly continue to play along.

The fundamental tension will not be because the AI companies are inherently bad actors, but because they will increasingly challenge the implicit authority of the state and most other existing power structures. That tension between digital technology firms and the status quo has existed for decades. In that sense, AI politics will be nothing new—just the same conflict, heightened.

The post is about many things, but here is how I’d describe it.

When, how and why should public policy analysts update their priors about truly transformative AI?

You should read his post to understand his take on the issue, and also to get a sense of why I describe what this post is about the way I just did.

What I want to talk about in today’s post is the following:

1. What are institutions, and why have they mattered?

2. Do they still matter as much today as they used to earlier?

3. If not, why not, and what has replaced them, and why?

4. Are AIs a complement for the work that you do? Or are they substitutes? If they’re somewhere along the spectrum, where are they located currently? Are they moving towards any one side, and if so, how fast?

5. Who will be in charge as the answer to pt.4 changes, and at ever increasing speeds?

6. TMKK?

---

What Are Institutions, And Why Have They Mattered?

Institutions, any student of economics will (should?) tell you, are the rules of the game. Institutions are actually three very separate, but also very inter-related things: formal rules, informal constraints and the characteristics of enforcing those constraints².

Here’s The Economist on what this actually means in practice, along with one example:

Like his contemporary Coase, Mr North understood that the transaction costs and frictions of the real world ruled out ideal solutions. He then set about exploring the ways in which societies created institutions to help ameliorate the problems created by these imperfections. As trade expanded in the Middle Ages, traders found themselves in need of ways to overcome all sorts of market shortcomings: to determine who was creditworthy, whose goods were of high or poor quality, which merchants were good partners for risky trade voyages, and so on. Different societies developed different institutional approaches to managing these problems: like common merchant codes and repositories of records tracking traders who ran afoul of it. Mr North studied the conditions under which such solutions might arise and be sustained (or lead to defections and collapse).

Institutions are persistent, he noted. Institutions developed as a response to one set of historical circumstances could inhibit development later on. Or encourage it; Mr North emphasised the role of Britain's Glorious Revolution in supporting British economic development. When Parliament conspired to replace King James II with William of Orange, that demonstrated the enhanced power of the legislature relative to the monarch, which in turn sent a credible signal that the monarch could be prevented in future from expropriating private wealth in a time of fiscal need. That constraint, he reckoned, was crucial in asserting the rule of law and the security of private property, clearing the way for Britain's economic revolution³.

If this is still mysterious and but-is-it-really-applicable-to-me for you, consider this:

1. Let’s say you are a young person in India, and you would like to work for Google. You are convinced that you will be able to generate more revenue for Google than whatever they will pay you, and you therefore think this to be a no-brainer proposal for both parties (you and Google).

2. How do you convince Google that your worldview is correct, and they should therefore hire you?

3. If only you could just land up at the Google offices and have a chat with them. That would be perfect (for you), and much less than perfect (for Google, because in such a world, you are not going to be the only person with such an idea).

4. So you might say well, get into a good college, study hard, crack the interview (assuming Google comes on campus, of course), and well, you’re in!

5. Now go and read “transaction costs and frictions of the real world ruled out ideal solutions. He then set about exploring the ways in which societies created institutions to help ameliorate the problems created by these imperfections” again.

Institutions, y’see? They’ve mattered.

---

Do They Still Matter As Much As They Used To?

Why do you need institutions? You need institutions because “transaction costs and frictions of the real world ruled out ideal solutions”. So if you live in a world where transaction costs and real world frictions go sharply down, then sure, you can say that your need for institutions isn’t quite as much as before.

Let’s take one more example, not related to getting a job at Google.

1. Let’s say you’re out for a drive in your car, and a cop stops you. You were stopped, he says, because you didn’t follow lane discipline. You, your co-passengers, everybody else on the road, and the cop knows that you were stopped because it was time to collect hafta.

2. If you play the “Tu jaanta nahi hai mera baap kaun hai” card, these are the “informal constraints” that are a part of the institution called traffic management in your city. The very fact that the cops are standing by the side of the road to impose traffic discipline collect hafta is itself an informal constraint, and you are responding with one of your own.

3. When you live in a city that uses cameras mounted on traffic signals to automatically send you a ticket to your registered home address, you have shifted to a world in which transaction costs have reduced, there are fewer frictions of the real world, and you therefore have not quite the same need for institutions in this case.

4. But if you insist on sticking to the getting a job at Google example, here’s homework for you.

So the answer to the question we’re talking about in this section depends on whether transaction costs and real world friction are on the way down, or up. If down, the current institutional set-up doesn’t matter quite as much.

Now, let’s talk about political institutions in your country.

---

If not, why not, and what has replaced them, and why?

“Risky thing to do no, to talk about the political institutions in my country?”, you might feel like asking me. You never know who will take offence to what, so why take a chance?

Well, true, and I quite agree with you, but I do hope you realize that I have no way of knowing who you are, and so I don’t know which country I’m talking about. The reason I can still speak meaningfully about you, your country and the political institutions within it, is because what I’m about to say applies globally.

The internet, you see, has significantly reduced transaction costs and real world frictions when it comes to feeling like you belong a political movement. You may not know the name of your local MLA, but you will know the name of your President/Prime Minister⁴. You can see them and hear from them on your social media of choice. You can see them and read from them, or about them, in your newspapers. They’ll be there on your phone, in your messages, in your inbox, on hoardings next to where you stay, and all around you in other ways you can’t imagine.

So ok, you might say, I get that price reductions when it comes to incessant communication implies that the supply of personalized communication from Dear Leader is going to go up. But why should the demand for this incessant and personalized communication be there, and that too from Dear Leader, as opposed to Entire-But-Not-Quite-So-Dear-Party?

Because we shut down in response to too much information. And make no mistake, we live in a world of too much information. So the winning strategy, in such a world, is to give dramatically simplified information, and have it come from but one source. You and I, we no longer have the bandwidth to listen to complicated stories and reach our own informed conclusions. Far easier to pick an easy-to-understand story from a trusted source, and stick with it.

So if you are a trusted source, and if you have an easy-to-understand story, you are likely to do better.

In a world of increasing complexity, the incentives are stacked in favor of a world led by individuals over institutions.

---

Are AIs Complements or Substitutes?

This gets us back to Dean Ball’s essay. Briefly put, he’s saying that there’s “really good AI”, which we shall say is complementary to human effort. The AI folks refer to this as human-in-the-loop, but sure, whatever works for you.

If/when AI transitions over to “transformative AI”, it will stop being complementary in nature. You may think this has already happened, or you may think it will happen in the future with x% probability. Whatever you think, how will the world adjust to this transition?

Remember, we’ve learnt that institutions arise because of transaction costs and real world frictions. If AI becomes transformative AI (becomes a substitute for us in everything we do, by becoming both cheaper and better at everything we do), then that will be a world with significant transaction costs and real world frictions⁵.

Which institutions will rise in response to this transition? Which institutions will go up in status? Which institutions will go down?

I don’t know about status, but Dean’s saying the quiet part out loud. Here’s the extract again:

The law enforcement of the internet will not be the government, because the government has no real sovereignty over the internet. The holder of sovereignty over the internet is the business enterprise, today companies like Apple, Google, Cloudflare, and increasingly, OpenAI and Anthropic. Other private entities will claim sovereignty of their own. The government will continue to pretend to have it, and the companies who actually have it will mostly continue to play along.

The fundamental tension will not be because the AI companies are inherently bad actors, but because they will increasingly challenge the implicit authority of the state and most other existing power structures. That tension between digital technology firms and the status quo has existed for decades. In that sense, AI politics will be nothing new—just the same conflict, heightened.

Dean is saying that the institutions governing the internet, and therefore those with the ability to Do Something in the age of the transition won’t be governments. It’ll be the internet firms.

Maybe. Who knows?

I have a more basic question. Whether it is the governments themselves, or whether it is the internet firms, the question remains:

Will AI be a complement for these institutions, or will it be a substitute?

---

Who Will Be In Charge When/If The Transition Starts?

Institutions (sovereign governments) as we know them? Or individuals? Or modern internet corporations?

Whatever the answer, to what extent will AI be a complement for each of them? To what extent will it have become a substitute?

How should we think about this problem, and how much time have we got?

This is a hard problem, and I do not pretend to have the answers to any of these questions. But yes, I have realized through writing this post that I need to think long and hard about institutions, individuals and intelligences.

---

TMKK?

1. Subscribe to Dean Ball’s Substack.

2. Read more about what economics has to say about institutions.

3. Read more about Truly Transformative AI.

4. Think long and hard about public policy and AI.

5. Learn how to update your priors about everything, especially sovereignty

1. I’m considering doing this with every post I write. Feedback (either way) would be appreciated

2. Source: https://heinonline.org/HOL/LandingPage?handle=hein.journals/catoj11&div=47&id=&page=

3. What happens when that constraint goes away? A friend of the current inhabitant of the British throne is about to find out.

4. Whoever is supposed to be top of the greasy pole in your country, basically

5. Understatement Pro Max Ultra