This post is a close-to-TLRD version of my Bledcom paper about the improbability of AI becoming human-like.

The exhaustive argumentation consists of 12.000 words, so either ask to mail you a PDF or wait for the paper to be published somewhere.

On Memetics and Meaning

Humans construct meaning through memetic fields—interactions of strategic and material memes. AI, which processes precepts without forming concepts, cannot host memes or generate its own subjectivity. While language models like ChatGPT assist us in meaning-making, they themselves lack intrinsic meaning.

On Agency and Morality

Agency emerges from the ability to choose among multiple memetic outputs. AI lacks this capacity, making deontological ethics inaccessible to it. Even if AI can simulate fear or decision-making (e.g., HAL in 2001), it cannot experience mortality or moral responsibility. Its decisions are inherently utilitarian and algorithmic.

On Consciousness and the Chinese Room

Searle’s Chinese Room illustrates that syntax is not semantics—AI may appear competent but lacks comprehension. Dennett’s idea that comprehension emerges from layered competence does not hold for AI because it lacks co-evolutionary embodiment with a phenotype.

On AI-Human Relationships

Public relations depend on intersubjectivity. Treating AI as a stakeholder or public presupposes agency and shared cultural ground. But AI cannot form genuine relationships, as it cannot participate in the intersubjective processes that define human culture and communication.

On Mortality, Culture, and Memes

Human culture evolves through memetic mortality. AI, being potentially immortal, cannot engage in culture as humans do. Culture arises from the finitude and fragility of individual minds. Memes require hosts who live and die. Immortal machines would lack the constraints necessary for cultural participation.

Conclusion

AI cannot know what it is like to be human. It may simulate intelligence, but it cannot attain wisdom. Until it develops the capacity for memetic infection, conceptualization, and mortality, AI will remain a powerful but non-human tool—useful, but never an agent. Understanding this distinction is essential for ethics, public discourse, and the development of AI systems.

Post Bledcom comment

After 33 years, Bledcom became an institution in the field of public relations (research and practice). From 120 – 150 participants each year, offering an option to discuss issues with such a broad spectrum of experts from all around the world that it is impossible not to find one that would challenge you.

While my belief that AI cannot become human-like remains as strong as it was before the conference, the evolution of self-replicating AI agents opens at least a need to rethink the evolution of non-biological self-replicating machines. The refutation of Dennett’s claim about the reducibility of consciousness to neural correlates remains, yet the replication of competence might result in the emergence of comprehension. What kind of comprehension might arise in AI agents remains open.

And, once again, AI cannot become an agent and, for that reason, also not a moral agent. Whatever AI evangelists and programmers say about the agency of AI, that agency is not like human or, for that matter, animal agency.

 

Some further readings:

Andrej Drapal. (2025). The Trauma and the Crisis. Management Studies, 13(1), 41–54. https://doi.org/10.17265/2328-2185/2025.01.005

Dennett, D. C. (2017). From Bacteria to Bach and Back: The Evolution of Minds (Reprint edition). W. W. Norton & Company.

Kant, I., Guyer, P., Wood, A. W., & Kant, I. (2009). Critique of pure reason (15. print). Cambridge University Press.

Mitchell, M. (2019). Artificial intelligence: A guide for thinking humans. Pelican Books.

Nagel, T. (1974). What is It Like to Be a Bat? Philosophical Review, 83(October), 435–450. https://doi.org/10.2307/2183914

Pinker, S. (2003). The blank slate: The modern denial of human nature (Nachdr.). Penguin.

Searle, J. R. (1980). Minds, brains, and programs. Behavioral and Brain Sciences, 3(3), 417–424. https://doi.org/10.1017/S0140525X00005756