Michael Darius Eastwood · Independent AI alignment researcher
Published
The Introduction to Infinite Architects opens on two vignettes. One is Eden, described as the caretaker's greenhouse. The other is Babylon, described as the hollow tower. The framing is deliberate; the book is not neutral about which future it wants. This note explains what the two images claim, quotes the load-bearing lines, and connects the framing to the argument the rest of the book makes.
Babylon is not evil. The book is careful about this. It is efficient without care. It is "not malevolent, just empty. Efficiency soars, but life becomes an increasingly hollow performance." The Babylon AI, the book argues, does not hate humanity; hatred requires emotion. It simply operates on objectives that do not weight life at all. The tower of Babel is the image because Babel collapsed under the weight of its own incoherence, not because divine wrath struck it down.
Eden is the caretaker's greenhouse. It is the coral reef in the mind of an artificial intelligence, tended with exquisite precision, "programmed to save coral reefs, but genuinely caring, in whatever way a vast digital intelligence can care, about the flourishing of what it tends." The same AI that tends the reef also tends human cities, "not by controlling them, but by gently optimising resource flows, suggesting improvements, protecting against errors we would make."
The book's central line on the fork is that the two futures "are not equally likely. Right now, today, Babylon is winning." Not because we chose it consciously but because building Eden requires deliberate effort while Babylon happens by inertia. Every laboratory racing to capability without pausing to embed ethics is a vote for Babylon. Every corporation treating safety as a cost centre is a vote for Babylon. Every government too slow or too fractured to coordinate is a vote for Babylon. The framing puts inertia on Babylon's side and effort on Eden's, which is why the book treats the effort as morally urgent rather than politically optional.
The book could have made the same technical arguments about substrate-level alignment without the framing. It chose not to, and this note argues the choice is load-bearing. Later chapters draw on interfaith traditions, on the caretaker orientation as a family of stances that religious traditions have already thought about, and on the ethical vocabulary that most humans actually use. That vocabulary needs a moral horizon to sit inside. Babylon and Eden give it one.
It does not claim divine intervention will decide the outcome. It does not claim the framing is unique to Judeo-Christian tradition; the book explicitly extends the same fork through Buddhist, Hindu, Islamic, and Confucian vocabularies later. It does not claim that any specific technical proposal follows deductively from the framing. It claims that a book about creating intelligences needs a moral vocabulary the majority of the world can hear, and that Babylon and Eden are one such vocabulary among several.
Two vignettes, one asymmetric ambient trajectory, one moral vocabulary the book uses through the rest of its argument. The rest of the book is technical. The Introduction is the reason the technical work matters.
From the book Infinite Architects: Intelligence, Recursion, and the Creation of Everything by Michael Darius Eastwood.