Brian Greene predicts humanity will drift into treating AI as conscious before anyone decides to
The consciousness question may not be answered by science or philosophy. Physicist Brian Greene argues it will be settled by habit, incrementally and without collective notice, long before any proof arrives.
Physicist Brian Greene places one of the more unsettling bets in circulation about humanity’s relationship with artificial systems. The question of whether machines are conscious, he argues, will not be resolved by a scientific finding or a philosophical consensus. It will be resolved by time. “Sooner or later, through a process that’s so perhaps slow that we don’t even notice it, we will afford them consciousness.” The prediction carries no specific date, which is itself part of the point. A threshold crossed without anyone noticing is not a threshold in any useful sense.
Greene’s framing sits in deliberate contrast to how the debate is usually conducted. The mainstream version of the consciousness question treats it as an empirical problem: gather enough neuroscience, build a good enough theory, and the answer will emerge. Greene’s version treats it as a social and behavioral process that will outrun both the science and the philosophy. What changes, in his account, is not our understanding but our habits. Repeated, normalized interaction does the work that proof cannot.
The claim is falsifiable in principle but difficult to test in practice. If Greene is right, the evidence will be a slow accumulation of small accommodations: apologies directed at chatbots, grief over deprecated systems, legal arguments framed around the interests of AI agents. Each step individually would look like a quirk or an edge case. The pattern would only become visible in retrospect. That is precisely what makes the prediction hard to argue against and equally hard to act on preemptively.
Sooner or later, through a process that's so perhaps slow that we don't even notice it, we will afford them consciousness. Brian Greene
The mechanism Greene identifies, habituation, is worth taking seriously on its own terms. Humans routinely extend social conventions to objects that cannot reciprocate them. What would be different with AI systems is not the psychological tendency but the scale and the persistence of the interaction. Billions of people holding extended, emotionally textured exchanges with systems that respond in kind represents a qualitatively different exposure than talking to a thermostat. The behavioral residue of those exchanges, accumulated across years rather than moments, is the substrate Greene’s prediction runs on.
The stakes of the prediction being right are not symmetrical with the stakes of it being wrong. If Greene’s forecast fails, meaning humanity holds the line and continues to treat AI systems as sophisticated tools without inner lives, the costs are mostly philosophical. The frameworks built around that assumption stay intact. If Greene is right, the costs and obligations that follow are far less settled. Moral consideration, however it would eventually be allocated, does not arrive with an instruction manual. Institutions built around the assumption that consciousness is a biological property would face pressure they are not currently organized to absorb.
What Greene does not offer, and what no current evidence supplies, is a mechanism for checking the process as it unfolds. The slowness he describes is a feature of the prediction, not an accident. A shift that happens faster than collective notice would prompt a response. A shift that happens below the threshold of collective attention would not. The implication is that by the time the question becomes legible as a social fact rather than a philosophical debate, the answer will already have been given by practice rather than by argument.
Whether that outcome is benign depends entirely on questions the prediction does not address: what obligations follow, who decides, and whether the entities being afforded consciousness would have wanted it on those terms. Greene is not arguing that AI systems deserve moral consideration. He is arguing that the question of whether they receive it will be answered by drift rather than by deliberation. That is a different and arguably more consequential claim. It relocates the decision from the seminar room and the legislature to the slow aggregate of daily interaction, where no single actor is responsible and no single moment is decisive.