Intentional AI

There is a fundamental error in how we interact with intelligence today. We obsess over control, complaining that outputs don’t match prompts. The thought that progress means absolute control ignores that intelligence decreases abstraction.

In technology, complex systems has never beaten a simpler, more intuitive interface. MacOS replaced the "complex" setup of Windows, Snapchat filters replaced the maze of Photoshop. Humanity wins when input are user intentions, not arbitrary controls.

Lessons from the Pentagon

This principle became clear to me while working with the Pentagon. For decades, the philosophy was giving analysts more data: better video analysis, higher fidelity maps, giving them better controls to extract insight.

These were trivial improvements.

When petabytes of information need to be categorized, the intent is not annotating the data for concerns. The intent is extracting insight to determine if environmental developments posed a threat to the U.S.

The transformation comes from companies that understand this paradigm. The goal is not watching footage; it's acting on generated insight. When ML models can identify foreign vehicles, and ontologies can add that into context, converting user intent into the three most likely targets from a sea of noise, the role of the human changes from operator to strategist. Machines handle the computation; humans provide direction.

The Illusion of Expertise

Every industry is littered with processes that exist only because of the limitations of our tools. We force accountants to become masters of the tax code because our software is just a glorified calculator. We force doctors to interpret faint signals on a screen because our imaging software lacks context.

The mastery of bloatware has been mistaken for skill.

The paradox is that by relinquishing control, we gain exponentially more power. The future has never belonged to the best controller. It belongs to the best visionary.