Modernization, Regulation and Costs

Benedict Evans chooses a timely title in his annual slide deck on technology. AI Eats the World.

In his presentation, he produces this slide:

The red circle on the right seems the most likely answer to his implied question. As that becomes real and fast, the range of human tasks that will be done by software and machines will grow, not linearly but exponentially.

Whether that will produce a highly centralized or decentralized world is a debating point this week. Matthew Harris argues the decentralization case, focusing on how new decentralized media personalities impacted the US election.

But this past election showed me beyond a doubt we've passed the tipping point. Decentralized media is officially more powerful than centralized media.

The definition of centralized and decentralized is changing, too.

Substack, often heralded as a decentralization success story, exemplifies the tension between independence and centralization. Writers left mainstream outlets for autonomy, but as John Gruber notes, Substack's uniform branding creates the illusion of decentralization while centralizing talent within its ecosystem. He takes his notes from a piece by Anil Dash - Don't call it a Substack

This tension between centralization and decentralization reflects a broader truth: while decentralization empowers individuals, centralized structures often provide the tools and scale to maximize their impact. Journalists and writers are leaving Main Street for their independent publications.

In Gruber's essay on Daring Fireball he rains on Substack's parade and accuses it of hijacking talent into its closed and centralized world. In Regarding - and, Well, Against - Substack, Gruber says:

Substack, very deliberately, has from the get-go tried to have it both ways. They say that publications on their platform are independent voices and brands. But they present them all as parts of Substack. They all look alike, and they all look like "Substack". I really don't get why any writer trying to establish themselves independently would farm out their own brand this way. It's the illusion of independence.

Now, I do see the point. But I don't object to it. Centralization is another word for building something. There is no built-in value for decentralized versus centralized. Indeed, most services to humans cannot be done in a decentralized way. Or at least they can be built better, faster, and cheaper if centralized.

There are many examples of great decentralized things. But in fast-moving technology, no decentralized application or service has ever been better than a centralized one. Protocols can be decentralized - like those that run the Internet - but products, services, or businesses are less so.

The rise of AI is likely to favor centralization, as building and scaling powerful AI products requires immense resources. Open-source, decentralized models will coexist, but the most impactful applications may emerge from centralized players. The accurate measure, however, isn't where innovation happens but whether it enriches human lives.

However, that will not stop decentralized models from being freely available as open source. Those open-source models will be used to build centralized products and services.

The central debate isn't about choosing sides in the centralized vs. decentralized battle but about outcomes. Whether through open-source collaboration or centralized innovation, the goal should remain clear: building tools and systems that help us live better, more fulfilling lives.

Population decline is not a frequent topic in That Was the Week, but Noah Smith's essay caught my attention. His view contrasts intriguingly with my good friend Phil Mullan's argument in The Myth of the Population Time Bomb. Together, they highlight a critical debate: is population decline a looming catastrophe or an exaggerated fear fueled by elite fatalism?

While Phil Mullan dismisses demographic doomerism as elite pessimism disconnected from economic realities, Noah Smith sees population decline as an undeniable problem requiring immediate remedies like birth-rate stabilization. Their approaches diverge sharply: Mullan focuses on productivity growth to offset declining populations, while Smith prioritizes active demographic interventions.

Today's demographic doomerism is not based on economics, let alone genuine concerns about the future. It is better understood as an expression of the fatalism and pessimism that now grips the imagination of our elites.

Noah looks at the facts of likely decline and assumes it has only bad consequences. So, he focuses on birth rate-boosting remedies. He states:

So how do we create a world where the human population neither explodes into overpopulation or trails off into an empty planet? The only way is active stabilization. We need ways to nudge fertility a little bit above the replacement rate when it's too low, or knock it down a little bit below the replacement rate when it's too high. We don't want to send fertility back to 5 or 7, but we need some way to be able to return it to 2.1 if we want to.

Phil focuses on productivity. He writes about the UK where the GDP per capita is around £33,000 compared to the USA's $85,000.

..a return to any decent productivity growth would overwhelm the arithmetical effect of population decline. With other assumptions staying the same, a productivity growth rate of just one per cent a year would see GDP per head double to over £65,000. With a return to the long-run productivity growth of two per cent, GDP per head would grow four-fold to about £140,000. Thus, even slightly raising productivity growth would swamp any income effects of the birth rate staying well below replacement levels.

This may be the old Malthus vs. Marx debate writ anew.

Another of this week's essays asks, "What Is the Earth's Carrying Capacity?" The results are fascinating. Scientists' estimates range from a quarter of the current population (at the low end) to as high as 128 billion at the high end. The bulk of the estimates lie between 8 billion and 64 billion. But a small number, not a tiny number, says we need half or even a quarter of the current numbers.

I side with the larger numbers. Humans seem capable of finding ways to live in all kinds of circumstances. The definition of resources changes as science discovers new things. The idea of fixed resources running out does not seem wise.

This brings us to the other critical factor influencing humanity's future: productivity. As AI continues to transform industries, it raises the question - can technological advancements make up for declining populations? Productivity and innovation are at the core of this week's debates, extending even to decentralization and media.

Essays of the Week