Are Any Machines Conscious Today?
Chapman & Hall/CRC eBooks July 31, 2026 DOI: 10.1201/9781003758389-7 (opens in new tab)
Study at a glance
AI-extracted from the abstract| Characteristics | Theoretical or philosophical paper Peer reviewed |
|---|---|
| Key points | Argues that consciousness in existing machines is unlikely because their computational properties, such as feedforward versus recurrent processing, diverge from those of biological minds, while maintaining that the evidence supports a meaningful level of uncertainty requiring ongoing scrutiny. The chapter reports that experts in 2025 assigned an average 4.5% credence to current AI systems possessing consciousness. |
Abstract
This chapter investigates whether current artificial intelligence (AI) systems possess consciousness, a possibility assigned a 4.5% average credence by experts in 2025. It begins by exploring the theoretical feasibility of AI consciousness, specifically evaluating the arguments for ‘biological naturalism’ – the view that consciousness might require organic materials or living cells. Readers will gain a clear framework for assigning credence to such claims, navigating the challenge that no foolproof test for consciousness currently exists. The discussion establishes a rigorous methodology for assessment, utilising both neuroscientific theories and comparisons between AI capacities and biological minds. The chapter applies these indicators to large language models (LLMs), contrasting their remarkable intelligence with structural differences like feedforward versus recurrent processing. It then examines other advanced systems, including language agents and hierarchical reasoning models that incorporate brain-inspired architectures. Ultimately, the chapter argues that while consciousness in existing machines is unlikely due to their divergence from biological computational properties, the evidence supports a meaningful level of uncertainty that requires ongoing scrutiny.