Core Concepts
Intelligence is a fundamental property of all systems that involves the differentiation, correlation, and integration of information to resolve uncertainty and achieve goals.
Abstract
The article presents the Theory of Intelligences (TIS), a unified framework for understanding intelligence across physical, biological, and artificial systems. The key ideas are:
- Intelligence involves two main capacities: solving (resolving local uncertainty) and planning (optimizing sequences of subgoals to achieve complex goals).
- Challenges to intelligence include goal difficulty (how system abilities compare to goal complexity) and surprisal (novelty of the goal).
- Intelligence can be expressed in informational units or in units relative to goal difficulty.
- The framework accounts for intelligence at multiple levels and scales, including how intelligence is transmitted within an individual's lifetime and evolves across populations.
- Mathematical models are developed to quantify solving, planning, and overall intelligence, incorporating concepts like information gain, path efficiency, and benchmarked ability.
- The theory argues that intelligence is a fundamental property of all systems, not just thinking entities, and that physical, biological, and artificial systems exhibit different forms and levels of intelligence.
Stats
"Intelligence is a human construct to represent the ability to achieve goals."
"Intelligence operates at many levels and scales and TIS distils these into a parsimonious macroscopic framework centered on solving, planning and their optimization to accomplish goals."
"Notably, intelligence can be expressed in informational units or in units relative to goal difficulty, the latter defined as complexity relative to system (individual or benchmarked) ability."
Quotes
"Intelligence is a fundamental property of all systems and is exhibited in a small number of distinct, distinguishing forms."
"The key advances of TIS are (1) the partitioning of intelligence into local uncertainty reduction ("solving") and global optimization ("planning"); (2) distinguishing challenges in the forms of goal difficulty and surprisal; (3) recognizing not only the core system, but extra-object spaces, including past sources, present proxies (i.e., any support that is not part of a system at its inception), environments, present and near-future transmission, and distant evolution."
"I do not discuss in any detail the many theories of intelligence nor the quantification of intelligence, the latter for which the recent overview by Hernández-Orallo (Hernández-Orallo, 2017) sets the stage for AI, but also yields insights into animal intelligence and humans."