The Alignment Problem Has an Architectural Assumption. QIS Breaks It.

The Assumption Baked Into Every Alignment Proposal Stuart Russell opened Human Compatible with a clean diagnosis: the problem isn't that AI systems are malicious. It's that we're building systems that optimize powerfully for objectives that don't fully capture what humans actually want. The AI does exactly what it was told to do. What it was told to do turns out not to be what we meant. The alignment literature - RLHF, constitutional AI, debate, amplification, scalable oversight, reward modeling - has generated a decade of sophisticated responses to this problem.