Architecture of Choice
AI alignment, moral architecture, and the oldest
engineering documentation in existence.
The Architecture of Choice
What the AI alignment problem and an ancient text have in common
The most important engineering problem of the twenty-first century may have been solved in the thirteenth century BCE. When you compare the architecture an engineer would design for training decentralized moral intelligence with the observable features of human existence, the parallels are striking.
The Speed Problem
Why AGI alignment can't afford the learning curve that humanity required
If the training of decentralized moral intelligence follows a characteristic curve, then replicating the architecture at machine speed, without the safety margins, creates a risk of extraordinary magnitude.
The Fractal
Why the same architecture repeats at every scale