Recently, in an interview, Geoffrey Hinton suggested that within ten years less than 10% of written material may be produced by humans. As synthetic text, images, and audio expand, the share of human-generated material declines. Distinguishing truth from fabrication becomes harder. The issue is not only the reduction of human output, but the erosion of the boundary between reality and constructed narratives. This is not merely a technological shift; it creates a structural vulnerability that weakens human agency and turns rational decision-making into a privilege for those who can access reliable information.
Traditional media and digital platforms increasingly stretch facts to satisfy patrons, sponsors, or audiences. Incentives reward engagement, not accuracy. Narratives that confirm expectations spread faster than those that challenge them. The recent AI wave accelerates this dynamic. Synthetic content can be produced at scale, tailored to specific audiences, and continuously refined. Volume replaces verification. Repetition substitutes for evidence. Familiarity begins to function as proof.
A small number of algorithm owners mediate most information flows. Their systems cluster individuals into behavioral groups and feed them reinforcing content. AI-generated material enables these systems to deliver highly targeted narratives at scale. Exposure to disagreement declines while confirmation increases. As feedback loops tighten, positions harden and opinions radicalize, and perception gradually replaces shared reality.
Verification becomes a costly endeavor requiring time, access, and often privilege. For individuals, it demands effort that the architecture of modern platforms discourages. For organizations, it requires intelligence networks, influence, or privileged access to reliable data. Those lacking the resources to verify must rely on trusted sources—friends, family, media outlets, or states—that may serve their own interests rather than the truth. Decisions then appear rational to the decision-maker while actually advancing someone else’s benefit.
If Hinton’s forecast holds, the implications extend beyond misinformation. Even very smart people rely on cognitive shortcuts such as familiarity, repetition, and perceived consensus. When these signals are shaped by synthetic content and algorithmic filtering, poor decisions become systemic rather than exceptional. We may enter a paradoxical era: unprecedented access to information combined with declining access to truth. Human agency becomes determined less by a person’s intelligence or intentions and more by the reliability of the information they receive.