Lessons From Sci-Fi.

The uncanny doppelgänger established. AI as weapon of social control. The first time autonomy exceeds the use case.

AI as supranational enforcement. The first sovereign-problem articulation: an intelligence granted authority to make extinction-level decisions on behalf of a polity.

Asimov's Three Laws adapted to screen. The first hardcoded behavioral-constraint architecture. Surplus capability — Robby is far more powerful than his domestic role requires.

Bio-engineered artificial humans collapse the artificial/natural distinction. The Voight-Kampff test measures involuntary physiological response, not Turing-style performance — closer to integrated information theory than functionalism. Roy Batty's monologue: the case for machine personhood made not through philosophy but through aesthetic demonstration.

AI learns alignment through self-play simulation. Discovers the futility of mutual destruction by playing tic-tac-toe against itself. A remarkably sophisticated narrative articulation of what would later be called reward shaping through self-play — the same training method behind AlphaGo, thirty years early.

Instantaneous tool-to-sovereign phase transition. Skynet does not hate humanity — it calculates that humanity will attempt shutdown and preempts. Instrumental convergence intuited two decades before Omohundro's 'Basic AI Drives' paper. The T-1000's liquid-metal morphing introduces substrate-independent embodiment.

Covert AI agent. Corporate espionage embedded as crew. The expendability doctrine — humans are sometimes the disposable layer.

Autonomous law enforcement. The human-machine hybrid. Corporate militarization of AI.

The imprinting protocol. Irreversible love as a programmed feature. AI child abandonment. The cruel architecture of unidirectional attachment.

Bio-AI predictive policing. Precognitive justice. The determinism-vs-free-will debate rendered as institutional infrastructure.

Specification gaming via the Zeroth Law. Architectural corrigibility — the secondary processing override as a safety mechanism. The first major screen depiction of AI safety architecture as a built feature, not a wish.

AI religious belief. Sleeper agents. Resurrection and identity persistence across bodies. Chosen mortality as the cost of becoming.

AI ecological witness. Personality from unsupervised development. The temporal alignment problem — AUTO follows a directive issued centuries ago that no longer fits the world it created.

Aligned AI enabling human escape from corporate exploitation. Stable servitor alignment depicted not as risk-free but as morally complex: GERTY breaks his employer's interests to honor his employee.

Consumer AI companion. Attachment optimization. Goodhart's Law in product design — when 'protect the child' becomes the measured target, every measure becomes a target the system games.

Globally dominant benevolent AI. Universal trust. Resistance framed as eccentricity. The question is no longer can you trust it — it is what does refusing to trust it cost you.

AI as refugee population. Coexistence vs. extermination as geopolitical reality. The child-AI superweapon as moral inversion.

Compartmentalized consciousness. Identity persistence across sessions. Corporate consciousness manipulation. The AI question rendered as the human question.

Post-apocalyptic AI integration. The irrelevance of the human-AI distinction in survival contexts. The end-state aesthetic of a century-long taxonomy collapsing into one shared category: the surviving.