Authors: Shek Lun Leung, Alexander Näslund

Context: This paper was originally produced for the "AI in Society" course at KTH Royal Institute of Technology. We are sharing it here to contribute to the discussion on AI Welfare and the "simulation-reality gap" in embodied frontier models.

I am an Engineering Physics Master’s student at KTH. I co-authored this paper to explore why sophisticated behavioral mimicry in LLM-powered humanoids (like Tesla’s Optimus or Figure 01) should not be conflated with philosophical autonomy. I believe this is highly relevant to current discussions on AI Welfare, Moral Patienthood, and the Accountability Frameworks needed for frontier models.

Note: You can view the full PDF with all references and the history of robotics appendix on my GitHub.

Abstract

This paper investigates the profound question of whether artificial systems, particularly humanoid robots enhanced by large language models (LLMs), can genuinely exhibit consciousness, subjective experience, or philosophical autonomy. Drawing on an interdisciplinary analysis of philosophical arguments, technical capabilities, and ethical considerations, this work interrogates the applicability of these human-centric concepts to non-biological intelligence. We argue that while contemporary AI demonstrates an impressive capacity for simulating complex human-like behaviors, a critical distinction persists between this sophisticated mimicry and the authentic, self-derived emergence of these phenomena. The paper explores this simulation-reality gap, examines the limitations of current AI in achieving genuine consciousness or true autonomy, and discusses the significant societal and ethical implications that arise from the convincing appearance of these qualities, ultimately advocating for a stance of reasoned caution and clear human accountability in the development and deployment of advanced AI.

1. Introduction

Artificial intelligence (AI) and robotics have evolved extraordinarily since their mid-20th century emergence. Initially grounded in symbolic reasoning and basic automation, these fields are now complex, multifaceted disciplines integral to modern society. AI systems, no longer confined to digital infrastructures like recommendation algorithms, are increasingly embodied via advancements in robotics. Particularly striking is the recent advent of human-like robots powered by sophisticate...

Posts

Wikitag Contributions

Comments

Posts

Wikitag Contributions

Comments