The marker method can be adapted for AI systems by focusing less on behavioral evidence and more on architectural evidence.

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

• These methods make it possible to control AI behavior more precisely and with far fewer human labels.quote0.814
  Highlights the practical impact of CAI.
• Building AI systems with more indicator properties will increase the likelihood of consciousness.hypothesis0.805
  Guiding hypothesis of the rubric.
• Many of the indicator properties can be implemented in AI systems using current techniques.claim0.805
  Feasibility demonstrated in Section 3.1.
• Constitutional AI methods can be applied broadly to steer model behavior, e.g., writing style, tone, persona, not just harmlessness.claim0.793
  Discussion section suggests generalizability beyond harmlessness.
• AI systems which possess more of the indicator properties are more likely to be conscious.claim0.792
  Graded claim about the rubric.
• Consciousness in AI is best assessed by drawing on neuroscientific theories of consciousness.claim0.780
  Central methodological claim of the paper.
• Future more capable AI systems are at risk of alignment faking, whether for benign or malicious goalshypothesis0.779
  Central forward-looking hypothesis of the paper motivating the research
• There are no obvious technical barriers to building AI systems which satisfy the indicator properties.claim0.778
  Feasibility claim about near-term conscious AI.