From Horsepower to Humanpower: Measuring the Cognitive Output of AI
When steam engines began replacing horses in the 18th century, James Watt needed a way to express their power in familiar terms. He coined “HORSE POWER (HP)” as a unit of mechanical output based on the equivalent work of the animals they displaced. This wasn’t just a technical measure; it was a symbolic pivot, capturing the shift from organic labor to industrial machinery.
Today, we’re witnessing a similar transformation — not of muscle, but of the mind.
Artificial intelligence systems are performing tasks once thought to require human cognition: reading scans, composing text, reviewing legal documents, generating strategies. These are not mechanical machines; they are cognitive machines. And as they begin to replace human intellectual labor, we need a new metric to understand their power: HUMAN POWER (also HP).
Just as horsepower (HP) compared engines to horses, humanpower (HP) can compare AI systems to human minds.
If an AI can write legal contracts at the speed and quality of ten junior attorneys, its output could be said to be 10 HP. A whole-body CT segmentation algorithm like our DAFS platform quantifying a 1200 slice CT scan in under 10 minutes, where it would take a human at an optimistic speed of 4 slices/hour, 300 hours to quantify with expert-level precision will be representing ~2000 HP in radioanatomical cognition. Unlike human cognitive output, which cannot be sustained beyond a few hours without rest and sleep, automation with thinking machines does not tire and can deliver consistent output continuously. Humanpower could serve as a bridge between machine capabilities and the social, economic, and ethical systems built around human cognitive work.
Philosophically, this reflects a profound shift. Horsepower measured mechanical effort. Humanpower would measure thinking effort.
We are quantifying not just what machines do, but how they think — or at least, how closely they approximate human reasoning. This calls back to Descartes’ dualism: if the body was mechanized in the industrial era, the mind is now undergoing its own abstraction into code and compute.
But to measure cognitive output is also to redefine it. What aspects of human thinking are we capturing — speed, accuracy, creativity, empathy?
What do we lose when we translate intellect into a scalar unit? Like horsepower before it, humanpower is not just a technical metric; it's a cultural metaphor for a species being redefined by its own inventions.
We stand at the threshold of a new age of automation. Humanpower may become the conceptual tool we need to navigate it — a way to reckon with the machines that don’t just do what we do, but think as we think.