The Human Alignment Problem

 

We are, in effect, building increasingly capable humans. We teach how to think through epistemology, how to perform through flow, and how to sustain performance through sleep, meditation, and recreation. This is our version of training more powerful systems. The result is clear: people become better at reasoning, creating, and executing. Capability increases.

But capability is not the same as alignment.

In the AI world, a system can be highly capable and still pursue the wrong objective. The same is true for humans. When we focus only on what to do, we produce individuals who are strong optimizers. They learn to reason well, perform well, and sustain effort. But we rarely specify what they should optimize for. In the absence of clear guidance, people default to optimizing for status, power, money, or short-term rewards. This is misalignment.

Once this happens, familiar failure modes appear. Objectives are mis-specified. People chase proxies like grades, likes, promotions, or wealth, mistaking them for real value. Goodhart’s law kicks in. Metrics become targets, and meaning is lost. Internally, there is conflict. Desire pulls in one direction, long-term goals in another. Anger overrides reason. Ego overrides cooperation. The person becomes a bundle of competing drives rather than a coherent agent.

Externally, the consequences are worse. Capable but misaligned humans can manipulate, exploit, and dominate. Intelligence amplifies impact, but not necessarily in a positive direction. Just like a misaligned AI, a misaligned human can cause harm more efficiently.

This is where how to do becomes critical. Traditions like Advaita and the Bhagavad Gita function as an alignment layer. They do not increase raw capability. Instead, they shape the objective function. Treat all equally reduces harmful goals. Truthfulness preserves clarity of thought. Humility keeps the system open to correction. Reducing desire, anger, and hatred stabilizes internal conflict. These principles do not weaken performance. They guide it.

With alignment, capability becomes a tool for reducing suffering. Without alignment, capability amplifies suffering. The difference is not in how smart or skilled a person is, but in what they are optimizing for.

This is the human version of the alignment problem. We do not just need more capable people. We need people whose capabilities are directed well. What to do builds powerful optimizers. How to do aligns what they optimize for.