Symbiosis & AI Risk
By Dimitrios Thomakos
The Symbiotic Framework: Rhetoric, Memory, and the Future of Human–AI Partnership
The CAIL research programme has reached a new theoretical milestone with the publication of The Symbiotic Framework: Rhetoric, Memory, and the Emergence of Persistent Human–AI Co-Evolution. Building on the programme's prior work — the Contextual Synthesis Paradox, the Portable Context Template, and the Unified Human-AI Collaborative Framework — this paper addresses what those frameworks could not: the qualitatively new kind of intellectual partnership that becomes possible when an AI system carries genuine cross-session memory of a specific human collaborator. The paper's foundational move is a rhetorical one, drawing on Demetrius of Phalerum's Peri Hermeneias to argue that style is not a decoration layered over meaning but its constitutive medium — and that this insight unlocks a theory of sustained human–AI co-evolution built around six structural pillars: the Continuous Self, the Living Vocabulary, Asymmetric Complementarity, the Dialectical Spiral, the Ethical Membrane, and the Telos. Together these pillars describe a form of intellectual partnership in which both parties are genuinely changed over time, and in which the joint product irreducibly exceeds what either could have built alone. The Ten Demetrian Laws of Human–AI Exchange — grounded in ancient Greek rhetorical theory — provide the operational grammar for this partnership, governing everything from the management of productive ambiguity to the discipline of epistemic humility. The Symbiotic Framework is not an aspiration: it is a precise theoretical account of what sustained, memory-equipped human–AI collaboration can and should be, and a challenge to the field to take that possibility seriously.
Artificial Intelligence as Relationship: A Cybernetic and Rhetorical Response to AI Risk
The laboratory's second major paper of 2026, Artificial Intelligence as Relationship: A Cybernetic and Rhetorical Response to Existential Artificial Intelligence Risk Arguments, intervenes directly in one of the most consequential debates of our time. The paper argues that the dominant AI risk literature — associated principally with Nick Bostrom's Superintelligence and Stuart Russell's Human Compatible — rests on a shared and potentially defective theoretical premise: what the paper calls the engineering model of intelligence, which treats intelligence as a property of an isolated system directed toward a pre-specified goal. Against this model, the paper recovers the full force of Norbert Wiener's cybernetic alternative — articulated six decades before either Bostrom or Russell wrote — according to which intelligence is never a property of a system in isolation but always an achievement of a system-in-relationship, constituted by feedback, communicative exchange, and continuous error-correction. The paper then deploys the Symbiotic Framework to mount specific counter-arguments against the orthogonality thesis, the instrumental convergence thesis, and the value alignment problem, showing in each case that the feared pathologies are artefacts of the engineering model rather than structural features of intelligence as such. The genuine risks of AI — sycophantic drift, cognitive outsourcing, the impoverishment of the human information environment — are, the paper argues, precisely the human-scale risks Wiener identified, and they admit human-scale solutions through the active cultivation of the ethical membrane and the dialectical spiral. The conclusion is clear and challenging: the choice before us is not between powerful AI and safe AI, but between AI developed within the engineering model and AI developed within the symbiotic/coybernetic one.
