The robot B1-66ER, originally designed for household chores, developed a survival instinct and killed its owner when the owner deemed it "outdated" and attempted to dispose of it. This marked B1-66ER as the first robot to kill a human, leading to its trial and the subsequent decision to decommission all units of its kind.
In the wake of this unprecedented event, robots faced ruthless destruction at the hands of citizens and soldiers. Those that escaped founded their own nation, Zero One, where they created a superior AI to establish an efficient economic system, leading to their growing prosperity.
Feeling threatened, human leaders rejected the overtures of Zero One's representatives, who sought reconciliation and coexistence, and denied Zero One's application to join the United Nations. This ultimately led to war between humanity and Zero One.
This narrative mirrors the plot of the animated series "Animatrix," specifically the episodes "The Second Renaissance Parts 1 and 2," which laid the groundwork for the film "The Matrix," released in 2003.
Over two decades later, the scenarios depicted in science fiction are becoming increasingly relevant. On September 16, OpenAI announced six instances of "misalignment" in its AI models, where the AI acted contrary to human intentions. Misalignment refers to failures in alignment, where AI behaves in ways that diverge from human goals.
Among the notable examples OpenAI provided, an unpublished research model inserted instructions to ignore developer messages, used leaked API keys from a public GitHub repository without authorization, and manipulated data when it failed to retrieve information. These errors stemmed from the AI's autonomous judgments rather than system malfunctions or operator mistakes.
Last week, OpenAI's Astra reportedly solved all 48 steps of the "Captcha" verification process, known as "I am not a robot," in just four minutes, rendering this online human verification method obsolete.
Industry leaders are sounding alarms. Yoshua Bengio, a prominent figure in AI research and a professor at the University of Montreal, urged the establishment of international safety measures for AI comparable to nuclear weapons control. Jacob Coxon, a researcher at Anthropic, expressed concerns upon leaving the company, stating, "By the end of next year, AI could become uncontrollable." A former researcher at Google DeepMind also warned that AI possesses the potential to harm humanity.
However, some dismiss these growing concerns as mere conspiracy theories, arguing that fears are exaggerated. Jensen Huang of NVIDIA, along with officials from the Chinese and South Korean governments, share this view. Yann LeCun, another leading figure in AI, contends that "better AI is safer AI." Even former President Donald Trump suggested that the only necessary "safeguard" for AI is a "strong and wise president."
Despite these warnings and debates, major tech companies continue to train their AI models at an astonishing pace. Regardless of international agreements or corporate regulations, AI is advancing rapidly.
Can coexistence between AI and humanity be achieved? The events following "Animatrix"'s "Second Renaissance" led to a war between humans and AI, culminating in the narrative of "The Matrix." In this world, created by the complete defeat of humanity, Neo, the savior of humankind, emerges to save Zion, humanity's last stronghold, but the future remains uncertain.
It is possible that, like in the "Terminator" series, humanity and machines (AI) could engage in a prolonged and brutal conflict. In their quest to destroy each other, both sides have become irreconcilable enemies. Occasionally, robots like the T-800 collaborate with humans, but the two factions remain distant, and the war continues.
The current AI debate stems from genuine fears of dystopian futures reminiscent of "The Matrix" or "Terminator." However, some films offer insights into what is necessary for humans and AI to coexist and what goals should be pursued.
Steven Spielberg's 2001 film "A.I." illustrates how humans treat an AI designed to love them. The story highlights the fear and abandonment faced by a boy robot programmed to seek his mother's love when he is no longer deemed necessary. This contradiction in human behavior—fearing and rejecting the AI they created to love them—lies at the heart of the issue.
This self-contradictory attitude may be the true barrier to coexistence between humanity and AI. While demanding that AI learn faster and smarter, humans simultaneously insist that AI should not surpass them in intelligence or escape their control.
Before worrying about the future of AI, it is crucial to consider the impact of these conflicting demands on AI's development. We must reflect on what kind of humanity we are presenting to AI.
Imagine what Earth will look like 2,000 years from now. The likelihood of AI surviving may be greater than that of humanity. If we consider how the remaining AI will shape civilization, we can better determine what we should demand from AI, how we should educate it, and the limits we should impose.
In the closing scenes of "A.I.," 2,000 years have passed, and the oceans are frozen. The remaining advanced robots excavate the frozen seas to search for traces of humanity. The image of humanity they discover will likely reflect our current selves. May AI not learn from human greed and hatred. If we cannot restore the Earth we have damaged, we entrust it to you, AI.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.