Technology's Centralizing vs Democratizing Forces
December 2024
Technologies can be classified by their tendency to either centralize or democratize power in society, with these effects driven by underlying economic and technological factors. This dynamic isn't new – technological innovations have always caused shifts in society that require rebalancing.
Some technologies inherently lead to centralization due to their fundamental characteristics. Nuclear power is a prime example: the massive economies of scale and rigorous safety requirements mean that only large reactors are viable. This necessitates concentrated funding sources and government oversight, making it a centralizing force despite its benefits.
In contrast, other technologies serve to democratize power and opportunity. Ride-sharing platforms like Uber and Lyft demonstrate this by distributing income-earning opportunities across a broad population, giving people more flexibility with their time and work. While these platforms do extract fees, competition between them helps keep these fees in check, with most of the excess value theoretically flowing to the drivers who compete in an open market.
The car versus train comparison further illustrates this dichotomy. The automobile was a democratizing technology, putting transportation power directly in individuals' hands. Trains, despite their efficiency, are inherently centralizing due to their massive infrastructure requirements, which typically demand government involvement. Both technologies provide clear benefits, but they differ in who controls the levers of power – either a distributed network of individuals or centralized authorities, whether governmental or corporate.
This brings us to one of today's most pressing questions: what will be the impact of AI? Like any transformative technology, its effects could go either way. One scenario is highly centralized: frontier AI models requiring enormous data centers and investment, controlled by a few powerful entities, potentially displacing many workers while concentrating benefits in these centralized hubs. The alternative is more democratic: if AI capabilities plateau at a level that allows for local deployment or small-cluster operation, the benefits could be widely distributed. In this scenario, workers across the economy would gain powerful tools to enhance their productivity, negotiate better compensation, and reclaim their time – all without the technology being centrally controlled by large corporations.
The key distinction often comes down to who captures the excess capital and value created by these technologies – whether it's distributed across many participants or concentrated in the hands of a few central players.
Dynamic Feedback Control in Government Policy
November 2024
Government policies can often be thought of as control systems, but many of them function as open-loop systems—static rules that don’t adjust based on outcomes. A clear example of this played out in India, where some states implemented policies to encourage smaller families while others did not. Over time, population shifts altered political representation, leading to unintended consequences. A more effective approach would involve designing policies with active feedback, dynamically adjusting in response to real-world conditions.
Taxes provide a useful example. Instead of a fixed fee, a step up in sophistication would be to structure taxes as a percentage of income, tying them to inflation or median earnings in an area. Taking it further, taxes could be linked to overall wealth or national debt levels, creating a system that adapts to economic conditions rather than remaining rigid. Similarly, congestion pricing in New York City currently functions as a flat fee, but a more effective system would make the cost proportional to real-time congestion levels, smoothing traffic flow rather than imposing a blanket charge.
Of course, dynamic policies come with potential pitfalls. A major one is opacity—if people can’t easily understand how the system works, it introduces confusion and delays in decision-making. But when implemented well, dynamic pricing mechanisms have been effective in balancing supply and demand across industries, and there’s no reason government interventions couldn’t function in a similar way.
Education policy offers another compelling case. A promising liberal idea is to provide more funding to underperforming schools rather than less, ensuring that all students receive a strong baseline education. But execution is key—if funding is simply tied to poor performance without safeguards, it could create perverse incentives for schools to underperform on purpose. A well-designed system would distribute resources dynamically while preventing exploitation.
Housing and zoning policies could benefit from the same approach. When housing prices spike in a city, regulatory barriers to building higher-density developments should be lowered, making it easier to meet demand. A land value tax could naturally enforce this principle by tying taxation to market value, ensuring land is used more efficiently without requiring constant intervention.
This broader idea differs from policies like minimum wage being indexed to inflation. The goal is not just to adjust a number in response to past data but to create an active control loop—something akin to a PID controller—where policies continuously respond to changing conditions in real time. The next logical step would be to model this in a simulated environment. For example, a population growth model could help determine what a stable demographic pyramid looks like and how policies might adapt to medical advances, immigration trends, or economic shocks.
Building such a simulation in Python, perhaps using Google Colab, could reveal fascinating insights about long-term policy stability. By treating governance as an adaptive system rather than a rigid set of rules, we could design smarter policies that evolve with society rather than lag behind it.
Statistically Accurate Character Generator
October 2024
I want to build a character generator that produces random characters, but with the correct weighted averages for a given population and time period. The idea is that randomness should still reflect real-world demographics. If I set it to Switzerland 200 years ago and ask for three random characters of a certain socioeconomic status, it should generate three Swiss people with a roughly even gender split. Simple enough.
But things get more complicated in the modern world. If I ask for ten random Americans, I’d want the generator to reflect the actual demographics of the U.S. in a way that’s statistically sound. Maybe six of them are white, with a 50/50 gender split, two are Black, and so on. Physical characteristics should also be randomized within realistic parameters—hair color, eye color, height, body type—without falling into stereotypes or biases.
A lot of this comes from the sense that representation in media often feels off in both directions. Hollywood sometimes overrepresents minorities in ways that feel forced, while also engaging in whitewashing, especially in settings like the Midwest where nonwhite characters are often ignored despite existing in real life. The goal here isn’t to push an agenda, just to create an accurate, useful tool that authors, screenwriters, or game designers can use to generate realistic character populations.
Ideally, this could even be integrated into storytelling. Think of something like a Mad Libs system where the core plot beats are laid out first, and character details are inserted afterward. Obviously, you can’t entirely separate identity from story—people’s backgrounds influence their experiences—but for certain types of storytelling, this approach could add flexibility and depth. A big inspiration for this is D&D campaigns, where NPCs are often randomly generated but still need to fit within the world’s logic.
Beyond race and gender, the generator should also account for factors like sexual orientation. If you’re creating a random group of characters, how many of them are likely to be gay? Bisexual? The tool should make sure these elements are represented naturally and proportionally, rather than leaning on outdated assumptions or arbitrary numbers.
The end goal is an online tool that lets users generate characters who feel real, with all the diversity and statistical probability of real human populations, while still embracing the unpredictability that makes randomness interesting.
Modular Cheap Bat Hibernacia
Summer 2024
WIP
Planetary Threat Timelines
Janurary 2025
WIP
3D Printed Home Design
Janurary 2025
WIP
Adaptive Supply and Demand Pricing
Janurary 2025
WIP
AI Archetypes
Janurary 2025
WIP
Anti-Matter Production
Janurary 2025
WIP
manual Delusion Generator
Janurary 2025
WIP