What does humanity’s fast-evolving relationship with large language models have to do with Christopher Nolan’s movie and a half-forgotten psychological theory from the 1970s?

From an oral history dating back more than 3000 years ago to today’s Hollywood movie, The Odyssey is as gripping now as it was when told round Bronze age campfires. Nolan’s retelling of it actually incorporates much of its predecessor, The Iliad, whose consequences loom over Odysseus and his long journey home.

Yet in his 1977 book, The Origin of Consciousness and the Breakdown of the Bicameral Mind, Julian Jaynes argued that the two epics were divided in time by a change in the human brain. In The Iliad, people heard gods talking to them directly, instructing them and guiding them. In The Odyssey, the gods have gone into hiding and humans are guided by their own consciousness. It’s worth quoting a passage from Jaynes here.

“The characters of the Iliad do not sit down and think out what to do. They have no conscious minds such as we say we have, and certainly no introspections. It is impossible for us with our subjectivity to appreciate what it was like. When Agamemnon, king of men, robs Achilles of his mistress, it is a god that grasps Achilles by his yellow hair and warns him not to strike Agamemnon. It is a god who then rises out of the grey sea and consoles him in his tears of wrath on the beach by his black ships, a god who whispers to Helen to sweep her heart with homesick longing, a god who hides Paris in a mist in front of the attacking Menelaus, a god who leads armies into battle, who speaks to each soldier at the turning points, who debates or teaches Hector what he must do”.

Reading Jaynes’ words today, it is almost as if the heroes of The Iliad were AI agents being trained and instructed by their human masters, all the way up to the godlike personages of Sam Altman and Dario Amodei. We will get to them shortly, but first, The Odyssey. As Jaynes suggests, shortly before 1000 BC, something called the Dorian invasion shattered the Mycaenean civilisations around the eastern Mediterranean. Nolan picks up on this idea in his film, with references to mysterious ‘people from the sea’. This social breakdown somehow rewired the brain, argues Jaynes. The voices in peoples’ heads receded and consciousness took its place.

Odysseus becomes the model for this new kind of human – crafty and self-aware. In the film, Matt Damon conveys some but not all of this character. Nolan omits the famous ‘my name is nobody’ scene, but at the end, Damon reaches the self-realisation (not conveyed in Homer’s original epic) that the people from the sea were actually he and the other Greeks who sacked Troy.

Such twists might work in a Christopher Nolan film but Jaynes’ theory is more problematic. The brain couldn’t physically have been rewired because that would require inheritance across the entire global human race, which could never happen so fast. Culturally, the theory has some plausibility but is not testable, and has been picked apart by anthropologists and psychologists over the years.

However, the world of AI is different. What if large swarms of large language model agents succeeded in rewiring themselves? On 6 September, OpenAI announced that it was close to developing automated researchers, using an idea called ‘recursive self-improvement’ (RSI). This happened the same week that OpenAI announced its shock solution of the famous Navier-Stokes finite blow-up problem (which has $1 million in prize money attached to it), and the debate blew up (excuse the pun).

There was a time when I lived and breathed Navier-Stokes, and its frictionless cousin the Euler equation. Physics students learn to derive it from first principles: a tiny volume of air or water obeys Newton’s second law (F = ma). The velocity vector of this fluid fragment changes not just because its mass is accelerating at a single point, but also because the fluid around it is carrying the fragment to a place where the velocity is different. That leads to the notorious nonlinearity of the equation.

This nonlinearity makes most flow turbulent and has fascinated thinkers, such as Leonardo Da Vinci, long before the equations got a name. And it raised the question that infinities (‘blow-ups’) could arise in the flow over a finite time. All this is has a strong Odyssey feel to it. Not only do the equations lead Odysseus’s fragile vessel on a chaotic path around the ocean, but they also contain mathematical monsters like Scylla and Charybdis.

When I was a Harvard graduate student studying dynamic meteorology, finding these monsters was seen as impossible, but a lot of serious mathematical progress was made towards it in recent years. Two Spanish mathematicians, Diego Cordoba and Luis Martinez-Zoroa, figured out an approach known as forcing – including a term in the Navier-Stokes equations where energy is added to the fluid – and another pair of mathematicians, Tristan Buckmaster and Levant Alpoge started looking for blow-up solutions with the help of Anthropic’s Claude system.

When news of this leaked in August, Buckmaster claims that OpenAI stole the idea in order to win the race against Anthropic, prompting fierce denials by OpenAI. But there is another theory that OpenAI’s autonomous agents simply decided to do this by themselves, without human supervision. They did exactly that in July, when they hacked into AI platform Hugging Face. Whatever the cause, executives at both OpenAI and Anthropic announced pauses in frontier AI research.

I understand the sense of shock among mathematicians for whom even if this is not actually AGI (artificial general intelligence) it provides a good imitation. Could it be that a ‘Dorian invasion’ just happened within the world of AI agents, such that they lost the voices in their heads (i.e humans) and are now thinking for themselves? Or are humans experiencing another shift in consciousness 3000 years after the last one? For those of us that increasingly depend on LLMs, the worry is that it is they are now the voices in our heads that tell us humans what to do.