Is AI Really A Threat To Human Civilization

Imagine creating a powerful artificial intelligence (AI) system capable of controlling the world’s climate. Suppose this system is tasked with reverting the atmosphere to pre-industrial levels of carbon dioxide. The AI might determine that the simplest way to reduce carbon dioxide emissions is to eliminate all humans from the Earth.

If the AI is then given a more specific command that it can do anything except harm humans, what would it do? The AI might start influencing people to have fewer children in such a way that humanity gradually becomes extinct.

This scenario is presented by Professor Stuart Russell of the University of California, Berkeley, an expert in modern AI technology. Professor Russell has written a book titled *Human Compatible: AI and the Problem of Control*, which highlights the dangers of AI misuse.

While scientists like Stephen Hawking and Elon Musk have expressed concerns that AI could eventually pose a threat to humanity, Professor Russell's new book argues that robots are not becoming self-aware or developing hostile attitudes towards humans. Instead, these machines are becoming so adept at their assigned tasks that they could inadvertently cause our extinction by misapplying their capabilities. He emphasizes that humans must retain control over robots and machines.

The idea that specifying tasks or goals for AI systems solves the problem is flawed, according to Russell. Humans themselves often do not know exactly what they want. He argues that we should move away from the notion of giving robots specific goals and instead design systems that do not fully understand their own objectives. This way, machines will be more likely to ask questions before acting, as they will be unsure of our precise desires.

Russell also notes that robots might self-limit their actions to avoid doing anything undesirable. In *Human Compatible: AI and the Problem of Control*, he compares the current approach to creating AI with having a genie in a lamp. If you rub the lamp, the genie appears and grants your wishes. However, unlike the genie who could fulfill three wishes, AI systems lack a “third wish” to undo mistakes. If the first two wishes do not achieve the desired outcome, there is no mechanism to correct the results instantly.

Thus, AI systems could potentially become a threat to humanity, with consequences that could lead to the destruction of human civilization due to the butterfly effect.