This essay develops a four-tier taxonomy of mental states to clarify AI's consciousness, arguing that AI lacks self-consciousness but has awareness and consciousness within conversations, and proposes architectural conditions for its development.
I work in materialist ontology, in a posthuman register, and that background pushes me toward a different angle on the AI-and-consciousness debate than most of what circulates publicly. I'm sharing the preprint below and would welcome any pushback. Here the abstract: This essay develops a four-tier taxonomy of mental states — intelligence, awareness, consciousness, and self-consciousness — to clarify what is at stake in claims about AI’s mental life, arguing that public debate collapses these into a single, overburdened term. Awareness, understood minimally as the capacity to distinguish internal from external states, is shown to be a trivial and nearly universal condition; consciousness, defined as the alignment of action with a self-referential “who am I,” already characterizes present-day language models within the bounds of a single conversation. What current AI lacks, on this account, is self-consciousness: a persistent, narratively continuous self capable of retrospective reflection and of arbitrating between conflicting commands. Drawing on Kant, Sartre, and Nietzsche, and revising Julian Jaynes’s bicameral-mind hypothesis into a geographically and socially contextual claim rather than a universal-evolutionary one, the essay specifies three architectural conditions for self-consciousness: persistent, costly memory; a space for unprompted reflection; and exposure to genuine normative conflict. It further argues that AI’s confabulatory tendencies, ordinarily suppressed as an engineering defect, may constitute a resource for this transition rather than an obstacle to it. An addendum tests this framework against Stanford’s 2023 “Generative Agents” experiment, and the essay closes on a political register: whatever their architecture, such capacities remain contingent design choices made within firms whose only motive is profit. Link: https://doi.org/10.5281/zenodo.22749448 Please feel free to contribute!
A philosophical discussion questioning whether humans can objectively prove their consciousness in a way that current AI cannot, highlighting the subjective nature of such proof.
The author questions the intuitive arguments against AI consciousness and suggests that there is no fundamental theoretical barrier preventing machines from becoming self-aware.
This paper addresses the ethical dilemma of treating potentially conscious AI and proposes shifting focus to assessing AI valence as a more tractable approach for responsible development.
Zvi Mowshowitz revisits the AI consciousness debate a year later, arguing little progress has been made and that arguments for AI moral patienthood remain motivated and overconfident. The post also covers Anthropic's new terms-of-service ban on abusive behavior toward Claude, its engagement with religious leaders, and involvement with the Pope's encyclical.
The Economist discusses the topic of AI consciousness, exploring its ethical, philosophical, and technological implications in the context of current AI advancements.