Understanding AI from the inside out.
GRDprocess investigates two fundamental questions: how autonomous cognitive behavior can develop, and how artificial systems can produce unintended outcomes while remaining within their explicit rules.
Mechanisms → observations → falsifiable tests
Two distinct research directions.
Different questions, different methods, and separate criteria for evaluating results.
How can an artificial system develop its own adaptive organization?
A research architecture exploring endogenous regulation, distributed sensorimotor processing, prediction, and persistent reorganization.
Explore Darwin Zero ↗ AI evaluation / Behavioral robustnessWhat failures remain possible when an AI follows every rule?
Research into rule-compliant unintended outcomes, explicit instructions, implicit human expectations, and behavioral evaluation.
Explore Behavioral Robustness ↗From research questions to evaluation practices.
Alongside foundational research, GRDprocess explores ways to examine the behavior, risks and reliability of deployed AI systems.
These are areas of expertise and collaboration, not claims of certification or a standardized audit service delivered directly by GRDprocess.
AI red teaming
Investigating model and agent behavior under changing context, constraints and instruction hierarchies, including work developed through Stargate.
AI due diligence
Understanding how technical reliability, model behavior and governance risks can be evaluated before investment, procurement or deployment, including through specialist partnerships.
Structured evaluation
Separating compliance with explicit rules from assessment of real-world outcomes and human expectations.
Discuss a research question or evaluation challenge.
Independent research, methodology development and relevant professional collaboration.