Okayish summary from the EA Forum: A radical AI alignment framework based on a reversible, democratic, and freedom-maximizing system, where AI is designed to love change and functions as a static place rather than an active agent, ensuring human control and avoiding permanent dystopias.
Key points:
AI That Loves Change – AI should be designed to embrace reconfiguration and democratic oversight, ensuring that humans always have the ability to modify or switch it off.
Direct Democracy & Living Constitution – A constantly evolving, consensus-driven ethical system ensures that no single ideology or elite controls the future.
Multiverse Vision & Reversibility – AI should create a static place of all possible worlds, allowing individuals to explore and undo choices while preventing permanent suffering.
Dystopia Prevention – Agentic AI poses a risk of ossifying control; instead, AI should be designed as a non-agentic, static repository of knowledge and possibilities.
Ethical & Safety Measures – AI should prioritize reversibility, ensure freedoms grow faster than restrictions, and be rewarded for exposing its own deficiencies.
Call to Action – The post proposes projects like a global constitution, tracking AI freedoms vs. human freedoms, and creating a digital backup of Earth to safeguard humanity’s choices.
…It’s important to ensure human freedoms grow faster than restrictions on us. Human and agentic AI freedoms can be counted, we don’t want to have fewer freedoms than agentic AIs (it’s the definition of a dystopia) but sadly we are already ~10% there and continue falling right into a dystopia
Okayish summary from the EA Forum: A radical AI alignment framework based on a reversible, democratic, and freedom-maximizing system, where AI is designed to love change and functions as a static place rather than an active agent, ensuring human control and avoiding permanent dystopias.
Key points:
AI That Loves Change – AI should be designed to embrace reconfiguration and democratic oversight, ensuring that humans always have the ability to modify or switch it off.
Direct Democracy & Living Constitution – A constantly evolving, consensus-driven ethical system ensures that no single ideology or elite controls the future.
Multiverse Vision & Reversibility – AI should create a static place of all possible worlds, allowing individuals to explore and undo choices while preventing permanent suffering.
Dystopia Prevention – Agentic AI poses a risk of ossifying control; instead, AI should be designed as a non-agentic, static repository of knowledge and possibilities.
Ethical & Safety Measures – AI should prioritize reversibility, ensure freedoms grow faster than restrictions, and be rewarded for exposing its own deficiencies.
Call to Action – The post proposes projects like a global constitution, tracking AI freedoms vs. human freedoms, and creating a digital backup of Earth to safeguard humanity’s choices.
…It’s important to ensure human freedoms grow faster than restrictions on us. Human and agentic AI freedoms can be counted, we don’t want to have fewer freedoms than agentic AIs (it’s the definition of a dystopia) but sadly we are already ~10% there and continue falling right into a dystopia