The Job Description Theory of Mind

In 1967, philosopher Hilary Putnam proposed something radical: mental states aren’t defined by what they’re made of, but by what they do. Pain is whatever receives damage signals as input, produces avoidance behaviour as output, and connects to other states like distress and memory in the right ways. The hardware is beside the point. Neurons, silicon, something else entirely — if it does the job, it has the mind.

This is functionalism. Its implications are simple to state and almost impossible to sit with. If minds are defined by causal roles rather than biological substrate, there’s no principled reason a sufficiently complex computer couldn’t be conscious. Not simulating consciousness. Having it.

Ned Block saw where this led and tried to break it. His 1978 thought experiment: imagine every person in China holding a radio, each simulating a single neuron, the entire population wired together to replicate a brain’s causal structure. Functionalism says this network has a mind — that there’s something it’s like to be the nation of China. Block meant this as a reductio ad absurdum. Many functionalists shrugged and accepted it.

Putnam didn’t survive contact with his own theory. By 1988, he’d rejected it, arguing that mental states are too context-dependent to be captured by functional roles alone. The father of functionalism became one of its sharpest critics.

But the framework won anyway. Every serious debate about AI consciousness — whether large language models “understand,” whether future systems deserve moral consideration — runs on functionalist assumptions. The question is never what is it made of? It’s always what can it do?

Liberating or terrifying, depending on how close you think we are to building something that does enough.