community
active
leiden_hybrid_concepts
label: haiku
community:leiden_hybrid_concepts-run4-c6-c0Aliveness as orthogonal to AI competence
Framework distinguishing perceived aliveness (roughness, contrast, boundary-awareness) from capability metrics in language models, examining how scale, polish, and user state affect felt presence.
15 members. Each node is clickable.
Loading graph…
Drawn from 5 sources
The papers/notes whose extracted claims & findings make up this cluster.
- 15-properties-of-aliveness-in-AI.md9 members
- koan-battery-section.md3 members
- 2026-05-09_briefing_for_ozero.md1 member
- Koan Battery: Measuring Reflective Mode Accessibility in AI1 member
- RESEARCH-VECTORS.md1 member
Bridges (3)
Other communities that share members with this one — cross-cutting threads or papers that sit at the seam between two themes.
Claims (14)
- A line must be drawn between aliveness and manipulation; casinos feel alive by some measures but are not ethically alive.
- Aesthetic response and boundary awareness are the most independent dimensions of interface aliveness.
- Aliveness in interfaces is not the same as competence; haiku-sized responses can feel more alive than expert ones.
- Aliveness may be relative to user state; an interface alive for a tired user may be exhausting for an alert one.
- Contrast is the property most violated by RLHF and most diagnostic of alive versus polished AI.
- Interface aliveness requires understanding the trust gradient—what the right depth-building is in a fresh conversation with an AI.
- Interpretability tools can reveal what 'feeling alive' looks like inside a neural network model.
- Mirror and Deathbed tests measure orthogonal facets of quality; what feels most alive differs from what merits preservation.
- Polish and smoothness reduce perceived aliveness; roughness signals real process over manufactured output.
- Smaller models produce more alive responses than larger ones in the same alignment family—roughness signals living process over manufactured polish.
- The right scale-set for measuring AI aliveness must be chosen: word/sentence/response, message/exchange/conversation, or lifetime relationship.
- The same 15 properties may not apply across modalities; aliveness in voice, text, and video may differ.
- What makes AI feel alive is almost nobody's question; aesthetics of aliveness is a core territory for About Blank.
- What you would preserve (Deathbed test) correlates with what you value and may serve as an alignment proxy.
Findings (1)
- Aliveness and competence come apart; Haiku outranks Opus in forced-choice aesthetic comparisons despite lower baseline.Alexander mirror method reveals smaller models produce rougher, more alive responses; competence (rubric) ≠ aliveness (aesthetic).