# Consciousness in Artificial Intelligence: Insights from the Science of Consciousness

**Authors:** Patrick Butlin, Robert Long, Eric Elmoznino, Yoshua Bengio, Jonathan Birch, Axel Constant, George Deane, Stephen M. Fleming, Chris Frith, Xu Ji, Ryota Kanai, Colin Klein, Grace Lindsay, Matthias Michel, Liad Mudrik, Megan A. K. Peters, Eric Schwitzgebel, Jonathan Simon, Rufin VanRullen  
**arXiv:** [2308.08708](https://arxiv.org/abs/2308.08708)  
**Date:** 2023-08-17  
**Summary page:** https://www.summarizepaper.com/en/paper/2308.08708v3/

## Key points

- The report explores the topic of whether current or near-term AI systems could be conscious
- The authors argue for a rigorous and empirically grounded approach to AI consciousness
- They assess existing AI systems based on neuroscientific theories of consciousness
- Several prominent scientific theories of consciousness are surveyed, including recurrent processing theory, global workspace theory, higher-order theories, predictive processing, and attention schema theory
- "Indicator properties" of consciousness are derived from these theories and used to evaluate AI systems
- No current AI systems are found to be conscious, but there are no technical barriers to building such systems in the future
- The report emphasizes the need for a comprehensive understanding of consciousness in order to accurately assess AI systems
- Ethical considerations associated with developing conscious AI are highlighted
- The report is authored by experts from various institutions and was developed through collaboration in workshops
- Patrick Butlin and Robert Long are joint first authors and corresponding authors for the report.

## Summary

This report explores the topic of whether current or near-term AI systems could be conscious, which is of scientific interest and increasing public concern. The authors argue for a rigorous and empirically grounded approach to AI consciousness by assessing existing AI systems in detail based on our best-supported neuroscientific theories of consciousness. They survey several prominent scientific theories of consciousness, including recurrent processing theory, global workspace theory, higher-order theories, predictive processing, and attention schema theory. From these theories, they derive "indicator properties" of consciousness that can be assessed in computational terms to evaluate AI systems for these properties. The authors use these indicator properties to assess several recent AI systems and discuss how future systems might implement them. Their analysis suggests that no current AI systems are conscious but also indicates that there are no obvious technical barriers to building AI systems that satisfy these indicators. The report emphasizes the need for a comprehensive understanding of consciousness in order to accurately assess AI systems and highlights the potential implications and ethical considerations associated with developing conscious AI. The report is authored by a group of experts from various institutions including University College London, University of Montreal, Araya Inc., Australian National University, New York University, Tel-Aviv University, University of California Irvine and Riverside, and Université de Toulouse. The authors collaborated in workshops to plan the report, develop ideas, and review drafts. Patrick Butlin and Robert Long are joint first authors and corresponding authors for the report. Overall, this report provides valuable insights into the assessment of AI consciousness based on neuroscientific theories and highlights the potential for future development of conscious AI systems while addressing ethical considerations surrounding this field.