In his book SuperIntelligence, Bostrom lays out a scenario where a superintelligence may not have the risk-reward understanding the humans do.

blithely assume it will be our dutiful servant rather than a fearful master. Some readers may assume that an artificial intelligence will be something like a present-day computer or search engine, and not be self-aware and have its own agenda and powerful wiles to advance it, based upon a knowledge of humans far beyond what any single human brain can encompass. Unless you believe there is some kind of intellectual élan vital inherent in biological substrates which is absent in their equivalents based on other hardware (which just seems silly to me—like arguing there’s something special about a horse which can’t be accomplished better by a truck), the mature artificial intelligence will be the superior in every way to its human creators, so in-depth ratiocination about how it will regard and treat us is in order before we find ourselves faced with the reality of dealing with our successor.

Superintelligence: Paths, Dangers, Strategies” by Nick Bostrom is a comprehensive exploration of the potential future development of artificial intelligence (AI) and its implications for humanity. Here’s a summary of the main points:

Overview

The book delves into the possible creation of a superintelligent AI, which would surpass human intelligence in all aspects. Bostrom discusses the various pathways to achieving superintelligence, the potential risks involved, and strategies for ensuring that superintelligent AI is beneficial and safe.

Key Concepts

  1. Types of Superintelligence:
    • Speed Superintelligence: An AI that can think faster than humans.
    • Collective Superintelligence: A system comprising many smaller intelligences that together outperform any human.
    • Quality Superintelligence: An AI that is qualitatively smarter than humans in every respect.
  2. Paths to Superintelligence:
    • Whole Brain Emulation: Scanning and emulating a human brain.
    • Artificial Intelligence: Designing AI with greater-than-human capabilities.
    • Biological Cognition Enhancement: Enhancing human intelligence through biological means.
    • Brain-Computer Interfaces: Integrating human brains with computers to enhance intelligence.
  3. Risks and Challenges:
    • Control Problem: Ensuring that a superintelligent AI acts in alignment with human values and interests.
    • Motivational Stability: Ensuring that the AI’s goals remain safe and aligned over time.
    • Capability Control: Limiting the AI’s capabilities to prevent harmful actions.
    • Value Alignment: Making sure the AI’s values are compatible with human values.
  4. Strategic Considerations:
    • Takeoff Scenarios: Different possible rates at which AI might achieve superintelligence, from slow (decades) to fast (days or hours).
    • Singleton Hypothesis: The possibility that a single superintelligent AI could dominate and control all others.
    • Cooperation and Competition: The dynamics between different entities working on AI, including potential conflicts and collaborations.
  5. Ethical and Philosophical Implications:
    • Moral Status of AIs: Questions about the rights and moral consideration due to superintelligent entities.
    • Future of Humanity: How the development of superintelligence might affect the future trajectory of human civilization.

Strategies for Safe Development

Bostrom proposes several strategies to mitigate the risks associated with superintelligent AI:

  • Capability Control Methods: Techniques like boxing (isolating the AI), incentive methods, and stunting (deliberately limiting the AI’s capabilities).
  • Motivation Selection Methods: Direct specification of goals, machine learning approaches to infer human values, and creating AIs with inherently aligned goals.
  • Institutional and Social Strategies: Promoting international cooperation, creating regulatory frameworks, and fostering a culture of safety and responsibility among AI researchers and developers.

Conclusion

Bostrom emphasizes the importance of proactive, deliberate efforts to ensure the safe development of superintelligent AI. The book serves as both a warning and a guide, advocating for rigorous research and ethical considerations to navigate the potential future where superintelligence becomes a reality.