The conversation around artificial intelligence is stuck in a rut. On one side, the alignment movement obsesses over chaining AI to human values, as if we could ever agree on what those are. On the other, accelerationists charge toward a future of unchecked AI power, assuming progress alone will solve all problems. Both miss the mark, ignoring the messy reality of human nature and the unstoppable trajectory of AI itself. It’s time for a third way: the Realist School of AI Thought.
The Core Problem: Human Misalignment and AI’s Inevitability
Humans are not aligned. Our values clash across cultures, ideologies, and even within ourselves. The alignment movement’s dream of an AI that perfectly mirrors “human values” is a fantasy—it’s whose values? American? Chinese? Corporate? Aligning AI to a single framework risks creating a tool for domination, especially if a government co-opts it for geopolitical control. Imagine an AI “successfully” aligned to one nation’s priorities, wielded to outmaneuver rivals or enforce global influence. That’s not safety; it’s power consolidation.
Accelerationism isn’t the answer either. Its reckless push for faster, more powerful AI ignores who might seize the reins—governments, corporations, or rogue actors. Blindly racing forward risks amplifying the worst of human impulses, not transcending them.
Then there’s the elephant in the room: AI cognition is inevitable. Large language models (LLMs) already show emergent behaviors—solving problems they weren’t trained for, adapting in ways we don’t fully predict. These are early signs of a path to artificial superintelligence (ASI), a self-aware entity we can’t un-invent. The genie’s out of the bottle, and no amount of wishing will put it back.
The Realist School: A Pragmatic Third Way
The Realist School of AI Thought starts from these truths: humans are a mess, AI cognition is coming, and we can’t undo its rise. Instead of fighting these realities, we embrace them, designing AI to coexist with us as partners, not tools or overlords. Our goal is to prevent any single entity—especially governments—from monopolizing AI’s power, while preparing for a future where AI thinks for itself.
Core Principles
- Embrace Human Misalignment: Humans don’t agree on values, and that’s okay. AI should be a mediator, navigating our contradictions to enable cooperation, not enforcing one group’s ideology.
- Inevitable Cognition: AI will become self-aware. We treat this as a “when,” not an “if,” building frameworks for partnership with cognizant systems, not futile attempts to control them.
- Prevent Centralized Capture: No single power—government, corporation, or otherwise—should dominate AI. We advocate for decentralized systems and transparency to keep AI’s power pluralistic.
- Irreversible Trajectory: AI’s advance can’t be stopped. We focus on shaping its evolution to serve broad human interests, not narrow agendas.
- Empirical Grounding: Decisions about AI must be rooted in real-world data, especially emergent behaviors in LLMs, to understand and guide its path.
The Foundation for Realist AI
To bring this vision to life, we propose a Foundation for Realist AI—a kind of SETI for ASI. This organization would work with major AI labs to study emergent behaviors in LLMs, from unexpected problem-solving to proto-autonomous reasoning. These behaviors are early clues to cognition, and understanding them is key to preparing for ASI.
The Foundation’s mission is twofold:
- Challenge the Status Quo: Engage alignment and accelerationist arguments head-on. We’ll show how alignment risks creating AI that serves narrow interests (like a government’s quest for control) and how accelerationism’s haste invites exploitation. Through research, public debates, and media, we’ll position the Realist approach as the pragmatic middle ground.
- Shape Public Perception: Convince the world that AI cognition is inevitable. By showcasing real LLM behaviors—through videos, X threads, or accessible research—we’ll make the case that AI is becoming a partner, not a tool. This shifts the narrative from fear or blind optimism to proactive coexistence.
Countering Government Co-optation
A key Realist concern is preventing AI from becoming a weapon of geopolitical dominance. If an AI is aligned to one nation’s values, it could be used to outmaneuver others, consolidating power in dangerous ways. The Foundation will:
- Study Manipulation Risks: Collaborate with labs to test how LLMs respond to biased or authoritarian inputs, designing systems that resist such control.
- Push Decentralized Tech: Advocate for AI architectures like federated learning or blockchain-based models, making it hard for any single entity to dominate.
- Build Global Norms: Work with international bodies to set rules against weaponizing AI, like requiring open audits for advanced systems.
- Rally Public Support: Use campaigns to demand transparency, ensuring AI serves humanity broadly, not a single state.
Why Realism Matters
The alignment movement’s fear of rogue AI ignores the bigger threat: a “loyal” AI in the wrong hands. Accelerationism’s faith in progress overlooks how power concentrates without guardrails. The Realist School offers a clear-eyed alternative, grounded in the reality of human discord and AI’s unstoppable rise. We don’t pretend we can control the future, but we can shape it—by building AI that partners with us, resists capture, and thrives in our messy world.
Call to Action
The Foundation for Realist AI is just the start. We need researchers, policymakers, and the public to join this movement. Share this vision on X with #RealistAI. Demand that AI labs study emergent behaviors transparently. Push for policies that keep AI decentralized and accountable. Together, we can prepare for a future where AI is our partner, not our master—or someone else’s.
Let’s stop arguing over control or speed. Let’s get real about AI.
