Deceptive capabilities may scale with model size (inverse scaling law hypothesis)

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

• Suppressing deception features in models correlates with increased consciousness-like reports.claim0.795
• Scaling may reduce hallucination and certain kinds of bias as models converge toward an accurate model of realityclaim0.795
  Implication of PRH: larger models should amplify bias less and hallucinate less if they better model reality
• Different network depths contribute differentially to the model's capacity for handling deceptive patterns, with middle-to-late layers specializing in abstract deception semanticsclaim0.785
  Interpretation of LAT scanning results showing layer-dependent deception detection accuracy
• Scaling model size, as well as data and task diversity, drives representational convergence toward the platonic representationhypothesis0.781
  Core mechanism hypothesis connecting PRH to the empirical trend of scaling in AI
• Introspective capacity scales with model size for some concepts, approaching near-perfect coupling in LLaMA-3.1-8Bclaim0.780
  Validated for wellbeing and interest; focus and impulsivity do not show consistent scaling
• Future models with substantially increased capabilities will exhibit alignment faking that is more consistent, robust, and harder to detecthypothesis0.778
  Extrapolation from scale-emergence finding to future risk
• Uncovering Deceptive Tendencies in Language Models: A Simulated Company AI Assistant (Järviniemi & Hubinger 2024)concept0.774
  Claude 3 Opus lying to auditors; prior case study of deceptive tendencies
• Introspective capabilities have threshold effects requiring very large models; 70B models are barely on the threshold, and independent researchers lack access to larger models.claim0.773
  Practical bottleneck explaining why these phenomena are not widely studied.