Creator
About the Architect
Stan Davis, Architect — founder of Lamp of Diogenes. He designs the epistemic architecture and cognitive regulation framework that defines the project’s conceptual and technical foundation.
Lamp of Diogenes develops the epistemic control layer modern AI is missing — starting with ANMA‑LoD, a synthetic neuromodulatory architecture for multi‑agent AGI.
Lamp of Diogenes starts from a simple premise: modern AI systems optimize, but they do not regulate themselves. They can predict, plan, and search — yet lack a global sense of how coherent or trustworthy their own beliefs are.
Epistemic Integrity (EI) is introduced as a scalar measure of epistemic health: a function of uncertainty, coherence, novelty, prediction error, and agent disagreement. EI is the foundation for all higher‑order cognitive regulation in ANMA‑LoD.
As systems scale, they become multi‑agent cognitive ecosystems: planners, critics, memory systems, world‑modelers, and retrieval agents operating concurrently. These subsystems hold different beliefs, update at different rates, and propose conflicting actions.
Without a unifying regulatory layer, the result is familiar: oscillatory behavior, runaway exploration, brittle generalization, catastrophic error cascades, and incoherent internal beliefs. This is not a training flaw — it is an architectural omission.
The Artificial Neuromodulator Architecture (ANMA) introduces a global control layer inspired by biological neuromodulation. ANMA defines synthetic analogs of dopamine, serotonin, acetylcholine, and norepinephrine, each a differentiable function of global cognitive statistics.
Integrated with Epistemic Integrity, these signals form a regulatory grammar for artificial cognition — shaping learning, exploration, attention, and anomaly response across all agents in the system.
Regulates learning rate, exploration, and meta‑plasticity. Increases with prediction error and novelty, and is modulated by EI.
Regulates planning horizon, patience, and volatility damping. Increases with EI and decreases with reward variance and disagreement.
Regulates attention, precision, and sensory weighting. Increases with signal‑to‑noise ratio, relevance, and EI.
Regulates anomaly response, high‑alert modes, and resets. Increases with anomaly magnitude and catastrophic error, and is suppressed by EI.
Every agent in the system receives the same global modulatory vector:
Mₜ = (sDAₜ, s5HTₜ, sAChₜ, sNEₜ, EIₜ)
Each agent interprets these signals through agent‑specific gain vectors, allowing explorers to increase plasticity, planners to extend horizon, memory systems to adjust retention, and world‑modelers to tune precision. One global signal, many heterogeneous effects — mirroring biological cognition.
Many catastrophic AI failures are epistemic failures: miscalibrated uncertainty, failure to detect anomalies, incoherent internal beliefs, brittle behavior under novelty, and inability to down‑regulate plasticity.
These are not solved by RLHF or scaling alone. They require epistemic regulation. ANMA‑LoD provides the missing layer: a global epistemic scalar, synthetic neuromodulators, and bounded modulation dynamics that stabilize multi‑agent cognition.
The ANMA‑LoD paper formalizes synthetic neuromodulation for multi‑agent AGI: defining Epistemic Integrity, neuromodulator equations, agent‑specific gain vectors, and a stability sketch under bounded modulation.
It also presents an illustrative multi‑agent gridworld simulation to characterize expected behavior under ANMA‑LoD, with improved exploration efficiency and reduced catastrophic failures compared to unregulated baselines.
Intelligence is not mere optimization. Intelligence is regulated optimization.
A mind without regulation is not a mind — it is an unbounded optimizer with no epistemic brakes. Lamp of Diogenes exists to design and formalize the regulatory mechanisms required for stable, coherent artificial minds.
ANMA‑LoD is a starting point, not an endpoint. Ongoing and future directions include:
Lamp of Diogenes explores the intersection of epistemic control, cognitive architecture, and artificial intelligence — developing the regulatory mechanisms required for stable, coherent artificial minds.
For collaboration or comments: contact@lampofdiogenes.com ·
Stan Davis, Architect — founder of Lamp of Diogenes. He designs the epistemic architecture and cognitive regulation framework that defines the project’s conceptual and technical foundation.