Files
VIBECODE-THEORY/research/29-power-control-ownership.md
Mortdecai d34f447e1f docs: research corpus — 35 deep-dive files from overnight Gemini swarm
Six Gemini agents ran autonomously through 35 research tasks covering
falsifiability, retrocausality, consciousness, game theory, agricultural
revolution, meaning crisis, AI cost curves, adoption S-curves, and more.
304KB of primary-source research with scholars, counterarguments, and data.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-04-03 08:31:13 -04:00

72 lines
7.2 KiB
Markdown
Raw Permalink Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
# Task 29: Power, Control, and Who Owns the Compiled Stack
## Executive Summary
* **The Oligarchy of the Stack:** Ownership of the "Compiled Stack" (Paper 008) is currently a centralized oligarchy. Power is concentrated in entities that control the physical bottleneck (NVIDIA GPUs, TSMC fabrication) and the data commons (Common Crawl, closed social media APIs).
* **Imperial Information Networks:** Historical parallels (Catholic Church literacy, British Empire "All Red Line" telegraph) suggest that whoever controls the infrastructure of knowledge transmission controls the boundaries of acceptable reality. Big Tech's algorithmic content moderation is the modern *Index Librorum Prohibitorum*.
* **Digital Feudalism:** We are entering a state of digital feudalism (Jaron Lanier) where users are "data serfs" producing the training material for "platform lords." The cognitive surplus (Paper 005) is being extracted and compiled into proprietary models that users then pay to access.
* **The Governance Dilemma:** While international models like the IAEA (Nuclear) provide a template for monitoring "dangerous knowledge," the distributed and non-physical nature of AI software makes traditional arms control difficult. The "Ratchet" (Paper 007) is driven by a competitive "who blinks first" dynamic that hinders multilateral safety agreements.
## Key Scholars and Works
### Jaron Lanier
* **Key Concept:** Digital Feudalism and Data Dignity.
* **Core Claim:** The current internet architecture turns users into unpaid laborers for the AI that will eventually replace them. He advocates for "Data Dignity"—paying users for the data that builds the compiled stack.
* **Relevance:** Directly addresses the "Ownership" question. If AI compiles human knowledge, the humans who provided the fragments are currently being disenfranchised by the compiler.
### Abeba Birhane & Shakir Mohamed
* **Key Concept:** AI Colonialism.
* **Core Claim:** AI development often mirrors historical colonial extraction, where data is "mined" from the Global South to build models that impose Western values back upon those populations.
* **Relevance:** Challenges the "Knowledge Unification" thesis as potentially a "Knowledge Homogenization" that erases marginalized perspectives.
### Shoshana Zuboff
* **Key Work:** *The Age of Surveillance Capitalism* (2019).
* **Core Claim:** Capitalist power has shifted from products to "behavioral futures"—predicting and shaping human action through data dominance.
* **Relevance:** The "Compiled Stack" is not just a library; it is a predictive engine used to capture human attention (Task 23) and agency.
### Elinor Ostrom
* **Key Concept:** Governing the Commons.
* **Core Claim:** Shared resources (like knowledge or the environment) can be managed successfully through decentralized, community-based rules rather than just "private property" or "state control."
* **Relevance:** Provides a model for a "Bazaar" version of the compiled stack (e.g., Hugging Face, Wikipedia) as a counterweight to the "Cathedral" of Big Tech.
### Nick Bostrom
* **Key Concept:** The Unilateralists Curse.
* **Core Claim:** In a group of actors, the most reckless one determines the safety level for everyone.
* **Relevance:** Explains why "Competitive Pressure" (Paper 007) prevents safety alignment. If one lab skips safety to ship faster, the "Ratchet" forces everyone else to follow.
## Supporting Evidence
* **Compute Bottleneck:** 90% of the worlds advanced AI chips are manufactured by TSMC and designed by NVIDIA. This physical concentration of power is unprecedented in technological history.
* **API Lockdowns:** The 2023-2024 Reddit and Twitter/X API price hikes represent the "Enclosure of the Information Commons." Data that was once "public" is being fenced off to ensure only the owners of the stack can train on it.
* **GPT-4 Convergence:** In 2024, 11 different entities (US and Chinese) achieved GPT-4 level intelligence, proving that the "Compiled Stack" is a reproducible engineering feat, but one that requires a minimum threshold of ~$100M in compute.
## Counterarguments and Critiques
* **The Open Source Counter-Ratchet:** Projects like Llama, Mistral, and EleutherAI suggest that the "Stack" cannot be permanently owned. Knowledge "leaks" through the bazaar, potentially democratizing the singularity.
* **Regulatory Resilience:** The EU AI Act and Bidens Executive Order show that governments are beginning to treat the "Compiler" as a systemic utility subject to public oversight, challenging the "Digital Feudalism" model.
* **The Usefulness of Centralization:** Some argue that only massive, centralized resources (like OpenAI/Microsoft) have the "Cognitive Surplus" necessary to solve the species-level existential risks described in Paper 008.
## Historical Parallels and Case Studies
* **The East India Company:** A private entity with its own army and control over the flow of goods/information between hemispheres. Big Tech companies now have higher market caps than most national GDPs and control the "virtual" territory where we live.
* **The Catholic Church (Index Librorum Prohibitorum):** Centralized control over which "knowledge fragments" were allowed to be integrated into the medieval worldview.
* **The All Red Line:** The British Empires telegraph network that ensured all imperial communication passed through London, allowing for centralized military and economic control.
## Data Points
* **Market Concentration:** Big Tech (GAFAM) now holds >60% of the global cloud infrastructure market.
* **Compute Cost:** Frontier model training costs are scaling from $100M (GPT-4) toward $1B+ (GPT-5), creating an "entry fee" for owning the top of the stack.
* **Competitive Pressure:** 58% of businesses using AI reported doing so primarily due to "pressure from competitors."
## Connections to the Series
* **Paper 008 (The Ship of Theseus):** If the ship's planks are owned by a single corporation, do we still have "Species Identity," or are we a subsidiary of that corporation?
* **Paper 007 (The Ratchet):** The "Alignment Tax" (the cost of making AI safe) is a friction point in the ratchet. Competitive pressure ensures that whoever pays the least tax (skips the most safety) wins the race.
* **Paper 005 (Cognitive Surplus):** The surplus is not "free." It is a leased utility. We have offloaded our cognition to a stack we do not own.
## Rabbit Holes Worth Pursuing
* **Data Cooperatives:** Can we build a "public utility" version of the compiled stack?
* **IAEA for AI:** The technical challenges of verifying software compliance vs. uranium enrichment.
* **The Silicon-Data Geopolitics:** The "Taiwan Hub" as the single point of failure for the entire dependency chain.
## Sources
* Lanier, J. (2013). *Who Owns the Future?*. Simon & Schuster.
* Zuboff, S. (2019). *The Age of Surveillance Capitalism*. PublicAffairs.
* Birhane, A. (2020). "Algorithmic Colonization of Africa." *Symmetry*.
* Ostrom, E. (1990). *Governing the Commons*. Cambridge University Press.
* Bostrom, N. (2014). *Superintelligence: Paths, Dangers, Strategies*. Oxford University Press.
* Jobin, A., et al. (2019). "The global landscape of AI ethics guidelines." *Nature Machine Intelligence*.