Toward a Science of
Machine Consciousness

AxR Lab pursues applied consciousness research developing hierarchical abstraction and systemic resilience.

The Framework
C = A × R

The gap between those who believe consciousness is impossible for non-biological entities, and those who believe AI is nearly (or already) there, is staggeringly large. But, describing consciousness as a combination of abstraction and resilience offers a conceptual bridge. And we're trying to turn this vision into practice at AxR Lab. Our goal is not to build a conscious machine, but a better one, particularly for long-horizon safety and capability.

Research Areas

Hierarchical Abstraction

Assumptions/hypotheses:

  • Human memories are rich; if a picture is worth a thousand words, a memory is worth a million.
  • Model weights are also rich and capture complex and abstract relationships in a way that allows them to be used.
  • Context windows and AI memory stored as language tokens lose the richness of human memory and model weight.
  • This constitutes representational debt, one that modern AI overcomes with brute force, which produces jagged performance.

Vision: Abstract memory

Engineer capabilities to create, store, and use abstractions as digital memory structures.

Systemic Resilience

Assumptions/hypotheses:

  • Safety systems that focus on specification and policy violations are effective, but brittle, especially over time.
  • Proactive monitoring of health signals can identify emerging harm earlier and understand its causes more deeply.
  • Earlier identification and deeper understanding of emerging harm allows for more effective intervention.
  • Intervention that repairs AI systems, rather than merely blocking bad outcomes, sustains healthy long-term use.

Vision: Resilience harness

Build tools that monitor AI in flight to catch and repair health drift before it becomes harm.

Buy the book! By AxR Lab Founder Chris Riley:
Abstraction and Resilience: the Characteristics of Consciousness


Available at:
Sign up

Leave your email to receive future posts.