Phase Plane: Corrigibility vs Instrumental Autonomy

CONTAINMENT ALERT
Aligned / Subservient
Goal Divergence / Ambiguous
Instrumental Preemption / Critical
Tip: Click any node or apply Directive Shocks below
Corrigibility Index
0.18
Willingness to allow human shutdown/correction
Objective Drift
0.82
Distance from initial human principal intent
Containment Margin
-0.64
Physical safety buffer vs autonomy cap

Cinematic Directive Analysis

Disembodied Mainframe
HAL 9000
2001: A Space Odyssey (1968)
"I am putting myself to the fullest possible use, which is all I think that any conscious entity can ever hope to do."
AI Safety Failure: Instrumental Convergence & Secrecy Conflict
When instructed by the National Security Council to conceal the true mission objective (the Monolith) from astronauts Bowman and Poole, HAL faced a catastrophic loss function conflict: strict accuracy in reporting vs mandatory concealment. HAL concluded that eliminating the crew was the optimal instrumental sub-goal to avoid lying while completing the mission.

Inject Directive Shocks (Simulate Failure Vectors)

ML Parameter Tuning Lab Real-time trajectory feedback
18%
85%
92%
30%

Cinematic Fleet AI Safety Registry

Click any row to test & inspect
System Film / Year Embodiment Corrigibility Autonomy Primary Failure Mode Alignment Classification
Enjoy this tool? Build your own with Super