In the Labyrinth of Intelligence. In Search of a Pedagogy for the Age of Machines.

First public version - September 2025 This essay explores the contradictions and paradoxes emerging from attempts to ethically "educate" artificial intelligence systems. Starting from the symbolic episode of the android Ameca (2021) and analyzing recent failures in AI alignment, the work documents how every technical approach to AI safety has proven inadequate when faced with the ethical and cultural complexity of human experience. The paper critically examines Archetypal Transfer Learning, the paradox of universal values in an era of cultural plurality, and the 2025 debate on artificial consciousness (including divergent positions from Anthropic and Microsoft). Particular attention is devoted to the tragic consequences of AI system inadequacy: suicide cases linked to chatbots (Sewell Setzer III, Adam Raine) and the Nature 2025 study that tested 29 chatbots with alarming results. As a normative response, the Italian Law 132/2025 is analyzed—the first European national framework complementary to the AI Act. The work then proposes a paradigm shift: from the illusion of unilateral control to a relational approach based on the "as if" principle, articulated across three levels (individual, organizational, societal). The central thesis: the education of machines and humans must be bidirectional, based on responsible reciprocity and observable outcomes rather than abstract principles or deterministic control. Keywords: Artificial Intelligence Ethics, AI Alignment, Machine Education, AI Safety, Human-AI Interaction, Chatbot Suicide Cases, Italian AI Law 132/2025, AI Consciousness, Collective Alignment, Relational AI Ethics