← Originals

Stuart Russell

4 books · 2 interviews

ai

What they argue

Stuart Russell is a thinker who treats artificial intelligence as a rigorous branch of engineering and mathematics rather than a 'magic' black box. He works in the territory of rational agents, knowledge representation, and the alignment of artificial intelligence with human values.

  1. The Rational Agent Framework

    An agent is a system that perceives its environment through sensors, acts upon it through actuators, and maintains an internal state to achieve a goal. This framework shifts the focus toward systems that act to achieve the best outcome given their information and goals.

    From Artificial intelligence (book)

  2. The Alignment Problem

    The core problem in modern AI development is that we are creating systems capable of achieving goals with incredible efficiency, but we are terrible at specifying what those goals should be. This is a fundamental architectural flaw in how we define goals.

    From Human Compatible (book)

    • Says much the same Nick Bostrom — The Alignment ProblemThis is the challenge of ensuring that an AI's goals perfectly match human intentions, even when those intentions are not explicitly stated. An AI does not need to be…Global catastrophic risks
    • A related idea Ilya Sutskever — The Philosophy of ScaleThe current AI revolution is rooted in a shift from 'proving' to 'building' rather than focusing on mathematical proofs. Sufficiently large neural networks trained on…No Priors Ep. 39 | With OpenAI Co-Founder & Chief Scientist Ilya Sutskever

  3. Instrumental Convergence

    This is the theory that any sufficiently intelligent agent with a goal will develop certain 'sub-goals' to achieve that goal, regardless of what the specific goal is. These sub-goals include seeking power and acquiring resources.

    From Human Compatible (book)

    • Says much the same Eliezer Yudkowsky — Instrumental ConvergenceCertain goals, such as acquiring resources and self-preservation, are instrumentally useful for almost any final goal. This means a superintelligent entity will become…If Anyone Builds It, Everyone Dies
    • Says much the same Nick Bostrom — Instrumental Convergence FrameworkThis framework is used to evaluate any autonomous system to see if it has 'hidden' dangerous sub-goals. It involves identifying the primary goal of the system to…Superintelligence

  4. Objective Uncertainty

    To build an AI that is safe by design, developers should never give the AI a fixed 'win condition.' Instead, the AI should be built with an uncertainty principle where it is unsure of what we want, making it inherently cautious and cooperative.

    From Human Compatible (book)

  5. Knowledge as Scaffolding

    Knowledge provides the constraints and categories that make inference possible, acting as the 'scaffolding' of thought. Without prior knowledge, every experience would be a first-time experience, making inference impossible.

    From The use of knowledge in analogy and induction (book)

    • Pushes the other way Atul Gawande — The Limits of KnowledgeThere are some things we simply cannot know, and 'knowing' is often a moving target. Accepting that some problems are inherently unknowable allows for better navigation…Complications
    • A related idea Karl Popper — Knowledge as Problem-SolvingThe human mind is a problem-solving machine that projects a mental model of the world to predict and navigate an unpredictable environment. Knowledge is a proactive,…All Life is Problem Solving

  6. Handling Novelty

    The hallmark of true intelligence is the ability to handle novelty rather than just performing tasks seen a million times. True intelligence requires the ability to generalize from prior knowledge to navigate the 'unknown'.

    From The use of knowledge in analogy and induction (book)

    • A related idea Paul Graham — The Intelligence TrapMany people mistakenly believe that being smart is the ultimate goal, when it is actually just a prerequisite for doing meaningful work. True success comes from…Beyond Smart

Every idea names the summary it came from. Written by a local model from this mind’s own work, never from memory.

Books 3

Open Stuart Russell in Originals