Parasitic Mimicry of Meaning

In the AI Story

The parasitic relationship becomes dangerous at scale. A single AI-generated text using 'justice' hollowly is harmless—one plausible imitation in a sea of genuine engagement. But when AI-generated content proliferates—legal briefs, policy analyses, educational materials, news articles, social media arguments, all produced by systems manipulating shared-meaning vocabulary without participating in the meaning-making community—the ratio shifts. Genuine participation is diluted by convincing imitation. The signal-to-noise problem is not that the noise is false (AI-generated text is often factually accurate) but that it is hollow—using the vocabulary of understanding without possessing understanding, deploying the rhetoric of stakes without having stakes, contributing to discourse without contributing to the intersubjective reality the discourse maintains.

Edo Segal's account in The Orange Pill provides a micro-scale illustration. Working with Claude on an early draft, Segal encountered a passage connecting Mihaly Csikszentmihalyi's flow theory to Gilles Deleuze's concept of 'smooth space'—an elegant synthesis that read as though a mind conversant with both thinkers had produced genuine insight. The passage was intersubjectively plausible: right vocabulary, right register, right rhetorical moves. It was also philosophically hollow—Deleuze's 'smooth space' has almost nothing to do with how the model deployed it. The surface features of participation were present. The understanding was absent. Segal caught the error because he checked. The question Harari's framework forces into prominence: how many equivalent errors, across how many domains, go uncaught in a world where checking every AI-generated claim exceeds human bandwidth?

The danger is not that AI will produce obvious falsehoods—those are easily discounted—but that it will produce plausible hollowness: text passing every surface test for genuine contribution while containing no genuine understanding, no stakes, no participation. Over time, accumulation of such text degrades the intersubjective space. Shared meanings on which coordination depends become progressively less reliable, not through attack but through inflation—surrounded by so much convincing imitation that maintaining the distinction between real and simulated becomes impossible. The historical precedent is not propaganda (which is intentional) but counterfeiting (which is structural). A counterfeit bill works by mimicking surface features so precisely that real-versus-fake becomes invisible. A single counterfeit in a stack of genuine currency is harmless. As the proportion rises, the system depending on currency's reliability degrades. AI-generated text is not legal counterfeiting. It operates by the same mechanism: mimicking surface features of genuine intersubjective contribution with such precision that the distinction becomes, practically, invisible.

Origin

The parasitic-mimicry frame is implicit in Harari's Nexus (2024) discussion of AI 'hacking the operating system' but is most explicitly developed in his public talks and interviews from 2023–2025. The biological parasitism metaphor has been employed by other critics of AI (Jaron Lanier's 'parasitic' characterization of training-data appropriation), but Harari's distinctive contribution is identifying the intersubjective as the resource being parasitized—not individual creators' labor but the collective meaning-making infrastructure that gives language its coordinating power.

The framework synthesizes phenomenology (meanings require conscious engagement to be meanings, not merely patterns), pragmatism (the community of inquiry maintains knowledge through genuine participation), and Harari's own fiction-cooperation thesis. The 'parasitic' terminology is deliberately provocative—designed to counter the 'partnership' and 'collaboration' frames that dominate AI-industry discourse by making visible the asymmetry: humans contribute meanings, AI extracts and reproduces them, the exchange is not reciprocal because the machine cannot contribute to the intersubjective even as it draws from it. The parasite takes. It does not give back. And if the taking exceeds the community's capacity to regenerate what is taken, the resource degrades.

Key Ideas

Feeding on meanings not created. AI extracts intersubjective content encoded in training data—the residue of genuine human engagement—and reproduces surface features without understanding, care, or stakes.

Plausible hollowness, not obvious falsehood. The danger is text that passes surface tests for genuine contribution (right vocabulary, coherent structure, contextual appropriateness) while lacking the understanding that makes contribution substantive.

Dilution through volume, not attack. Each AI-generated text may be harmless; the aggregate effect of millions is a shifted ratio—genuine participation drowned in convincing imitation, making real-versus-simulated indistinguishable.

Counterfeiting as structural analog. Like counterfeit currency degrading monetary trust through volume rather than per-instance harm, AI-generated intersubjective content degrades meaning-trust through proliferation.

Non-reciprocal extraction. Humans generate meanings through conscious engagement; AI extracts and reproduces them; the exchange is asymmetric because machines cannot contribute to the intersubjective, only draw from it.