Alignment

Work on AI alignment, value learning, and ensuring AI systems behave as intended.

Posts

• The Mechanics of Agency: Maximal Theoretical Agent - Explores Maximum Theoretical Agent (MTA)—opposite extreme from MVA.
• True Neutral - Uses D&D alignment grid as metaphor for philosophical positions.
• Comparing Value Systems - Proposes using cosine similarity from vector mathematics to quantify alignment between value systems
• Ghosts in the Machine - A critical response to Mustafa Suleyman’s warning about “Seemingly Conscious AI” (SCAI), this essay
• Structural Alignment - This post marks a pivotal shift in Axio’s alignment framework, arguing that traditional goal-based a
• The Axionic Agency Sequence - This meta-post serves as a comprehensive roadmap and overview of the entire Axionic Agency research
• Axions as a Type of Agency - Introduces “Axion” as precise noun naming constitutive structural configuration of reflective agency
• Axionic Agency — Interlude III - Documents project’s pivot from “Axionic Alignment” to “Axionic Agency.” Records shift in understandi
• Alignment Under Uncertainty - Addresses alignment when outcomes, values, and agent capabilities are uncertain.
• Alignment Beyond Epistemics - Argues alignment discourse over-focuses on epistemic problems (what AI knows/believes/understands) w
• The Load-Bearing Parts of Agency - This post explains results from Axionic Agency VIII.6, which used ablation methodology to identify n
• You Can’t Align a Hurricane - This accessible essay uses the hurricane-vs-nation analogy to explain why AI agency matters for safe