About the author
Gunnar Zarncke
Author of Towards Superintelligence Alignment.
Gunnar Zarncke is the founder and Managing Director ofaintelope, a German nonprofit focused on AI alignment research; previously CTO/CISO building a security-first systems and teams for a fintech startup. Gunnar is a systems architect specializing in quickly evolving systems. Over more than a decade, he has led architecture and engineering across startups and larger orgs, including building a fintech platform and team as CTO, owning security as CISO, and driving an evidence-based security-first engineering culture. Gunnar is interested in the study of brain-like AI safety approaches, foundations of agency, collaborative AI, and AI safety in real-world systems.
About this project
Most of the website text is AI-generated from the main project and book material. The essays are hand-authored orientation, not synced chapter text. On chapter pages, section and subsection headings can show authorship chips matching the book. Enable with Notes (bottom corner).
Links
- LessWrong
- Twitter / X
- Substack
- GitHub
Schedule a call
aintelope
aintelope is a German nonprofit focused on AI alignment research. Its public tagline: AGI research inspired by neuroscience. A better future for all sentient beings.
Publications
- Towards Superintelligence Alignment: Boundaries, Values, and Correction(PDF)
- Parameters of Metacognition — The Anesthesia Patient
- Value Learning Needs a Low-Dimensional Bottleneck
- The Friendly Telepath Problems
- Is the Invisible Hand an Agent?
- Handles Before Interventions: Access-Model UAD and the Embedded Semantics of Agency Tests(PDF)
- Recoverability of Smoothed Agent Boundaries in Unsupervised Agent Discovery(PDF)
- Stealth–Capability Bounds for Multi-Resolution Unsupervised Agent Discovery(PDF)
- AI Safety Interventions(PDF) · LessWrong
- A Formalization of Acausal Trade on Top of Unsupervised Agent Discovery(PDF)
- Attractor Basins of Cooperation, Privacy, and Parasite Persistence(PDF)
- Bitwise Intelligence: A Blanket-Information Measure of Competence(PDF)
- Consciousness and Agency Backbone: A Minimal Operational Stack(PDF)
- Construction Without Understanding: Successor Agents and the Limits of Copying(PDF)
- Endogenized Intentional Stance: Predictive Compression and Goal-Rational Priors(PDF)
- Foundations of Unsupervised Agent Discovery in Raw Dynamical Systems(PDF)
- Loop–Hub–Value Model: From Free-Energy Loops to Intrinsic Values(PDF)
- Mistral Large 2 (123B) seems to exhibit alignment faking
- Preference vs. Capability: Value-Conditioned Prediction and Control Channels(PDF)
- Status Regulation as Free-Energy Loops(PDF)
- Stratification of Free-Energy-Minimising Loops in the Vertebrate Brain(PDF)
- Thou art rainbow: Consciousness as a Self-Referential Physical Process
- Unexpected Conscious Entities
- Unit of Caring: Architecture, Suffering, and Cross-Scale Aggregation(PDF)
- Unsupervised Agent Discovery
- Value Bundle Drift(PDF)
- When "HDMI-1" Lies To You
- Alignment Attractor: Executive Summary and Platform Framing
- aintelope project update
- Brain-like AGI project "aintelope"
- Estimating Brain-Equivalent Compute from Image Recognition Algorithms
- Summary of my Participation in the Good Judgment Project
- When does technological enhancement feel natural and acceptable?