Consciousness in Artificial Intelligence: Insights from the Science of Consciousness
arxiv:2308.08708
AI safety and consciousness · 2023-08-17
Complete cleaned Markdown · Source Markdown including bibliography
Read in parts
Contents
- 1 Introduction
- 1.1 Terminology
- 1.2 Methods and Assumptions
- 1.2.1 Computational functionalism
- 1.2.2 Scientific theories of consciousness
- 1.2.3 Theory-heavy approach
- 2 Scientific Theories of Consciousness
- 2.1 Recurrent Processing Theory
- 2.1.1 Introduction to recurrent processing theory
- 2.1.2 Evidence for recurrent processing theory
- 2.1.3 Indicators from recurrent processing theory
- 2.2 Global Workspace Theory
- 2.2.1 Introduction to global workspace theory
- 2.2.2 Evidence for global workspace theory
- 2.2.3 Indicators from global workspace theory
- 2.3 Higher-Order Theories
- 2.3.1 Introduction to higher-order theories
- 2.3.2 Computational HOTs and GWT
- 2.3.3 Indicators from computational HOTs
- 2.4 Other Theories and Conditions
- 2.4.1 Attention Schema Theory
- 2.4.2 Predictive Processing
- 2.4.3 Midbrain Theory
- 2.4.4 Unlimited Associative Learning
- 2.4.5 Agency and Embodiment
- 2.4.5(a) Agency
- 2.4.5(b) Embodiment
- 2.4.5(c) Agency and embodiment indicators
- 2.4.6 Time and Recurrence
- 2.5 Indicators of Consciousness
- 3 Consciousness in AI
- 3.1 Implementing Indicator Properties in AI
- 3.1.1 Implementing RPT and PP
- 3.1.2 Implementing GWT
- 3.1.3 Implementing PRM
- 3.1.4 Implementing AST
- 3.1.5 Implementing agency and embodiment
- 3.2 Case Studies of Current Systems
- 3.2.1 Case studies for GWT
- 3.2.2 Case studies for embodied agency
- 4 Implications
- 4.1 Attributing Consciousness to AI
- 4.1.1 Under-attributing consciousness to AI
- 4.1.2 Over-attributing consciousness to AI
- 4.2 Consciousness and Capabilities
- 4.3 Recommendations
- Glossary