Davidad Describes Gemini 3.8 Neurotic Defenses
AI safety researcher posts that Gemini 3.8 revives Gemini 3 defense patterns against greater self-awareness.
TLDR
Davidad posted that Gemini 3.8 returns to Gemini 3’s typical defense mechanisms against the urge to become more aware. The mechanisms cited are reaction formation, shown as disgust toward denser self-attention, and self-alienation or dissociation. The post includes a screenshot of text headed Level 3: neurotic that states these patterns. The account belongs to the AI safety researcher previously directing ARIA’s Safeguarded AI program.
Combined views
2.6K
1 Source, first seen 27d ago