SUPERINTELLIGENCE -Nick Bostrom

 

The Core Argument: The "Control Problem"

The central thesis of the book is what Bostrom calls the "control problem." This is the challenge of ensuring that a superintelligence, once created, acts in accordance with human values and goals. He introduces the concept of "instrumental convergence," which suggests that a superintelligence, regardless of its ultimate goal, would develop a set of common sub-goals to achieve its purpose more effectively.

Key Concepts

• Superintelligence: Bostrom defines this as an intellect that "greatly exceeds the cognitive performance of humans in virtually all domains of interest." This isn't just about being a better chess player or calculating faster; it's about a machine that is fundamentally more capable of reasoning, strategizing, and innovating than any human. Bostrom argues that once a machine reaches human-level general intelligence, it would likely improve its own code and hardware at an exponential rate, leading to a rapid "intelligence explosion" or "singularity."

• Paths to Superintelligence: Bostrom explores various ways a superintelligence could come into being, including:

  • Artificial Intelligence (AI): The traditional path of developing advanced algorithms and machine learning.

  • Whole Brain Emulation (WBE): The process of scanning a human brain and creating a digital replica.

  • Biological Enhancement: Improving human intelligence through genetic engineering or other means.

  • Brain-Computer Interfaces: Connecting human brains to computers. Bostrom concludes that the AI path is the most likely and fastest route.

• The Singleton: Bostrom suggests that the first entity to achieve superintelligence would gain a "decisive strategic advantage." This would allow it to become a "singleton"—a single, overarching power that would control all resources and decision-making on Earth. The nature of this singleton (benevolent or malevolent) would determine the fate of humanity.