Borges in the AI Loop: Free Will and Algorithmic Destiny

The modern iteration of the age-old question of who is truly in control, the user or the system, is being amplified by artificial intelligence. When you binge-watch a show based on Netflix's recommendation, you signal your approval, even without explicit feedback. This dynamic is central to the "human in the loop" concept in AI, which examines the extent of human involvement in automated systems, from movie recommendations to ChatGPT.

At one end of the spectrum, AI operates autonomously; at the other, it follows human commands precisely. The core question is how much of our interaction with AI, the movies we watch, the ideas we adopt, the news we consume, is shaped by the system versus our own choices. This process, known as reinforcement learning from human feedback, creates a cycle where AI learns from our reactions, which are themselves influenced by the AI, reminiscent of the spider-man pointing meme.

This idea is not new; it echoes long-standing tensions between formal authority and real power in organizations, and the philosophical debate on free will. AI, however, makes this problem specific and urgent. Forty years after his death, Jorge Luis Borges offers a relevant perspective in his 1960 poem "Ajedrez" (Chess). He writes of chess pieces unaware that a player governs their destiny, bound by "adamantine rigor" that constrains their will. Borges poses the chilling question: "God moves the player, and he, the piece. What God behind God begins the plot / of dust and time and dream and agony?"

Borges offers no answer, perhaps because the question itself is paramount. Recommendation and automated forecasting systems function like a dog trained to please its owner. The dog learns to act a certain way based on cues, anticipating a reward. Borges' question extends beyond the dog's actions to the complex interplay that shapes both the animal's behavior and its owner's. Similarly, AI forecasts influence user actions just as users shape the AI's development, creating a continuous feedback loop.