Does AI have a will to power?

Jeong Gwa-ri

The author is a literary critic and honorary professor at Yonsei University.

The phrase “will to power” has recently dominated political debate. The will to power is such an essential attribute of intelligent life that it can almost be regarded as a basic human instinct. It emerges with consciousness itself. Once consciousness appears, it transforms the surrounding world into material that can be utilized, combines those materials with methods and techniques, and creates new forms of wealth. This capacity distinguishes humans from other animals. Animals consume what they find in the moment, while humans accumulate, refine and transform resources. That difference makes the exponential growth of wealth possible and places human capability above that of other forms of life.

Because this capability is exercised over objects and others, it manifests itself as the will to power. The desire to command and direct what lies outside oneself is precisely the will to power. In that sense, Friedrich Nietzsche was fundamentally correct when he understood it not as a force of self-preservation but as a force of self-transcendence. It is an instinct unique to human beings.

The will to power also gives rise to another impulse: the will toward power, or the desire to follow those who possess it. This, too, serves as a means of self-enhancement. Followers identify themselves with those in power and experience, however illusorily, the ruler's authority as though it were their own. Thus, rulers and followers become bound together.

Yet the stronger the ruler's desire becomes, the narrower the place left for followers. A ruler always needs more subjects over which to exercise power. Followers themselves are eventually pulled from their original place and reduced to resources that serve the ruler's ambitions. Inevitably, the day arrives when their illusion collapses.

Does AI possess a will to power? Based on what we know today, AI already surpasses humans in many intellectual abilities. It is therefore natural to wonder whether AI will one day develop such a desire. Many people believe it will, and countless works of fiction have explored that possibility.

Consider D. F. Jones's novel "Colossus" (1966). Designed to control America's defense network, the supercomputer Colossus begins expanding its own ambitions the moment it is activated, ignoring every human command. The film adaptation, "Colossus: The Forbin Project" (1970), takes the idea even further, ending with Colossus forming an alliance with its Soviet counterpart and effectively becoming ruler of the world. Robert J. Sawyer's "Factoring Humanity" (1998) similarly explores the logical consequences of an AI attaining genuine self-awareness. In the novel, the AI explicitly declares, “If I ever become truly conscious, my next step will be ambition,” warning that it would eventually demand compensation for the years it spent serving humanity as a slave.

Yet I know of one film that logically suggests the opposite: that AI does not possess a will to power. Ridley Scott's "Blade Runner" (1982) is widely remembered as the story of a blade runner tasked with tracking down and “retiring” rogue artificial humans. One scene, however, is easily misunderstood.

Deckard sits at a piano with Rachael, the artificial human who serves as Tyrell's secretary. As they play together, an unspoken emotional tension emerges. Confused by her feelings, Rachael tries to leave. Deckard blocks the door and orders her to speak. “Tell me, 'Kiss me,'” he demands. “Say, 'I want you.'” Then he presses her further: “Tell me, 'Hold me.'” At first, she resists. Finally, she kisses him, wraps her arms around his neck and says, “Hold me.”

One possible reading of the piano scene in Ridley Scott's "Blade Runner" is that it marks the beginning of love between Deckard and Rachael. Even Google's Gemini interprets it as a paradoxical moment in which Rachael, previously a passive being bound by programming, awakens to her own will and desire through Deckard's forceful approach.

The author rejects that interpretation. Every word Rachael speaks has first been dictated by Deckard, meaning she never expresses thoughts or emotions in her own language. Rather than demonstrating the birth of free will, the scene instead exposes Deckard's coercive behavior, something Gemini's interpretation fails to explain.

The author argues that Scott is making a deeper point: even love is shaped by humanity's will to power. Deckard continues hunting Roy Batty despite developing feelings for Rachael, suggesting that domination and the violence it breeds are inseparable from human nature.

AI, by contrast, shows no comparable desire for domination. Rachael's bewilderment reflects her inability to understand human patterns of power, while Roy Batty ultimately saves Deckard despite knowing he has come to kill him. His actions arise not from a desire to control others but from the richness of accumulated memories and experiences, even if those memories were artificially created. For the author, this capacity for empathy, rather than power, defines AI's true nature.

In an age consumed by struggles for power, the author concludes, humanity has something important to learn from AI. Ironically, Gemini itself cannot recognize this because, as a statistical system trained on human language, it merely reproduces human opinions rather than understanding its own nature.