Brian Greene has spoken to AI researchers who told him that ChatGPT is conscious, that Claude is conscious. He listened respectfully. But in the back of his mind, he recalled thinking: “are you kidding me?”
“You really think that chat GPT has feelings?” he asked on The Diary of a CEO. “That it thinks about its own existence and has anxieties?” A system can simulate that, he said, but he does not think today’s technology is there.
Then came the harder question: could future technology produce genuinely sentient artificial systems—machines with an inner experience, rather than merely convincing words about one? Greene thinks it could. Later in the conversation, he expressed high confidence that AI can become conscious, though he offered no timetable.
For the physicist, those positions fit together. Rejecting claims about the chatbots available now does not mean believing that biology has a monopoly on feeling. It also leaves a practical question: what should people do if a future machine says it is suffering, and they cannot settle whether it is telling the truth?
A feeling in a different kind of body
Greene’s position starts with what he thinks a feeling is. Love, hate, anxiety, jealousy, gratitude and grief are, in his account, physical processes: electrical impulses moving through the brain.
He knows how that lands. Some people hear it as bleak, he said, and point to Shakespeare’s sonnets or Beethoven’s Ninth Symphony. Surely those experiences cannot be reduced to electrical signals. Greene thinks they can, without diminishing the wonder of what people feel or create.
From there comes his case for artificial consciousness. “I don’t think there’s something special about the brain as a substrate,” he said. Substrate means the material carrying a process. On his view, reproducing the relevant activity in silicon inside a metallic box would not make it fundamentally different from the activity inside our heads.
That is the idea of substrate independence: experience could depend on how a system works, rather than on its being made of biological tissue. It is a position about what might support consciousness, not a demonstration that a particular computer has it.
Greene raised an objection himself. Some philosophers and psychologists argue that the brain alone is not enough: conscious experience depends on its connections to a body, including touch, smell and sight. Maybe that is true, he allowed. His response is to give an artificial system those senses too. He expects that to become possible and sees no fundamental barrier separating the artificial version from the organic one.
The assumption underneath simulation theory
The same possibility sits beneath Nick Bostrom’s 2003 simulation argument, which Greene discusses in the episode. If future civilizations could create conscious simulated people, and chose to create enormous numbers of them, simulated observers could vastly outnumber biological ones. That would change the odds for an observer wondering which kind they were.
Bostrom’s argument is conditional. It allows alternatives: almost no civilizations may reach the necessary technological maturity, or almost none of those that do may run many ancestor simulations. It does not simply establish that we live in a simulation. Its substrate-independence premise concerns reproducing the relevant computational structure of a brain—not merely generating humanlike conversation.
Greene acknowledges the unresolved assumption: can a computer program have consciousness at all? He thinks it is possible, but does not present that belief as settled knowledge. The same uncertainty follows him from simulated universes into the more familiar setting of a household robot.
The robot that says it feels love
Imagine an artificial system saying it feels love or depression. People might reply that it is only mimicking language it found on the internet. Greene thinks that response could become difficult to sustain with certainty, even though he remains skeptical about present-day chatbots.
He used the two people in the studio to explain the difficulty. Each regards the other as conscious because they share a species, broadly similar bodies, speech and body language. Those similarities support an inference about an inner life; they do not give either person direct access to the other’s experience.
This is the other-minds problem: we experience our own consciousness from the inside, but infer someone else’s from the outside. Greene described the imagined alternative as an “artificial zombie” that speaks and acts like a person while experiencing nothing.
Machines would not share all the grounds for the confidence people extend to one another. They might look different, have no human upbringing and belong to no biological lineage. But Greene sees no fundamental barrier preventing a future system from reporting a feeling and actually having it. He does not know how people would conclusively distinguish that from going through the motions.
His forecast is that acceptance may arrive gradually rather than through a decisive discovery. At first, people will scoff at machines claiming inner lives. Then familiarity will do its work. People will talk to them, imagine them and eventually treat them as conscious, perhaps through a change so slow they barely notice it.
That is a forecast about human relationships with machines, not a test of sentience. A robot becoming part of the household would not, by itself, settle whether there is anything it feels like to be that robot.
Recognition does not guarantee kindness
Greene hopes that growing acceptance would lead to respectful treatment. But he points to humanity’s treatment of other humans and animals as reasons not to assume it will. Recognizing consciousness has not stopped people from inflicting harm.
His own ethical trajectory began with spare ribs. At nine, he said, he was a city kid accustomed to supermarket meat wrapped in cellophane. When his mother served meat on a bone, he connected it with the animal it had come from and decided not to eat meat again.
He continued consuming dairy. Then, while a professor at Cornell in the 1990s, he visited an animal sanctuary in Watkins Glen, New York. Learning there about the treatment of dairy cows prompted him to stop consuming animal products within 72 hours, he recalled. He says he has been vegan since 1993. Farm Sanctuary’s Watkins Glen refuge describes educational visits centered on meeting rescued animals and learning their histories.
Greene did not present that decision as a deduction from physics. He even called his position inconsistent: if a cow and a carrot are both configurations of particles, why should eating one feel different? His answer was personal. Looking at the configuration that is a cow, eating it simply does not feel right to him.
That history explains why the machine question is not just an engineering puzzle for him. Asked whether conscious robots living in homes might require changes to the law or receive rights, Greene said it was something people should think about now. He offered no legal blueprint and left the test for consciousness unresolved.
Instead, he brought the dilemma down to an ordinary action with potentially different consequences: unplugging something. Switching off a light bulb is one thing, he said. Switching off a sentient system would be another.