Artificial General Intelligence (AGI) – brief





    Definition

      A hypothetical stage of AI where a system matches or exceeds human cognitive abilities across any task.

      Represents the ultimate goal of AI: replicating human‑level intelligence in software or machines.

    Historical Context

      Explored since the early days of AI research; term “AGI” popularized in 2007 by Ben Goertzel.

      Contrasts with narrow AI, which excels only in specific domains (e.g., chess, medical diagnostics).

    Key Distinctions

      AGI vs. Strong AI – Strong AI focuses on consciousness; AGI focuses on broad task performance, though the concepts overlap.

      AGI vs. Artificial Superintelligence (ASI) – ASI exceeds human ability in virtually all tasks; AGI is human‑level across tasks.

    Common Proposed Definitions

      Turing Test (human‑like behavior).

      Brain‑inspired models.

      Human‑level performance on cognitive tasks.

      Ability to learn new tasks with limited data.

      Economically valuable work.

      Flexible, general capabilities (e.g., Gary Marcus’s benchmark tasks).

      “Artificial Capable Intelligence” (ACI) – AI that can achieve complex, real‑world goals.

    Current State of LLMs

      Large language models (LLMs) show broad abilities but lack true AGI: they miss common sense, real‑world interaction, persistent memory, and reliable performance across all tasks.

    Technological Approaches to AGI

      Brain emulation – Replicating human brain processes in software.

      Novel architectures – New AI designs distinct from both brain and narrow AI models.

      Integrative approach – Combining narrow AI tools (LLMs, vision models, RL agents) under a central “agent” to achieve broader capability.

    Timeline Outlook

      Most experts predict a 50 % chance of human‑level AGI by the end of the century; some surveys now suggest a median estimate around 2047.

    FAQs

      AGI vs. AI: AI is any decision‑making system; AGI is AI with human‑level, general intelligence.

      Agentic AI: Enables autonomous, multi‑step actions within defined domains; not the same as AGI.

      Foundation models: Useful building blocks but not AGI on their own.

      Self‑awareness: No consensus; some view consciousness as required, others do not.

      How to know when AGI is achieved: No single benchmark; various performance tests (e.g., ARC‑AGI) exist but remain subjective.


Comments

Popular posts from this blog

IBM goes big to keep quantum computers cold

IBM - avances en mantener adecuadamente fríos computadores cuánticos.

Estructura multinivel de los sistemas de agentes de IA