The capacity for unlimited associative learning is not a good indicator for consciousness in AI.

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

• Consciousness in AI is best assessed by drawing on neuroscientific theories of consciousness.claim0.848
  Central methodological claim of the paper.
• Unlimited associative learning capacity (UAL-capacity)concept0.817
  Open-ended capacity for associative learning proposed as a transition marker for consciousness.
• Building AI systems with more indicator properties will increase the likelihood of consciousness.hypothesis0.810
  Guiding hypothesis of the rubric.
• AI systems which possess more of the indicator properties are more likely to be conscious.claim0.805
  Graded claim about the rubric.
• Both under-attributing and over-attributing consciousness to AI carry significant risks.claim0.800
  Risk summary.
• We take our principal contributions in this report to be: 1. Showing that the assessment of consciousness in AI is scientifically tractable ... 2. Proposing a rubric for assessing consciousness in AI ... 3. Providing initial evidence that many of the indicator properties can be implemented in AI systems using current techniques ...quote0.794
  Summary of contributions.
• Can we develop better behavioural tests for consciousness in AI that are difficult to game?question0.792
  Open question from Box 4.
• Tests of performance on specific tasks, including language modeling, are insufficient for determining consciousness statusclaim0.788
  Systems directly optimized for output can produce it without the prerequisite processes for conscious experience; simplest explanation for LLM consciousness reports is pattern matching