Chapter 01 · Section III · 10 min read
Philosophy of AI
The Turing test, Searle's Chinese room, and why "can a machine think?" is the wrong question to organise a country's policy around.
Philosophy of AI is older than computers. In 1950, Alan Turing — who had spent the war breaking codes — asked a deceptively simple question: can machines think? He then immediately said this question is too vague to answer, and replaced it with something more practical: can a machine imitate a human well enough that another human cannot tell?
That swap — from thinking to imitating — is the founding move of modern AI. The history of the field is, in many ways, the history of getting better and better at imitation.
The Turing test, briefly
You sit at a terminal and exchange typed messages with two correspondents. One is a person, one is a machine. You ask anything you like. If, after a serious effort, you can’t tell which is which, the machine has passed.
Modern language models pass this test in many domains. They fail in others. Whether they truly understand what they are saying is — by Turing’s design — outside the test. Turing’s bet was that this question wouldn’t matter very much.
The Chinese Room
John Searle, in 1980, disagreed. Imagine a man locked in a room with a rule-book for manipulating Chinese characters. Someone slips a Chinese question under the door. The man, who speaks no Chinese, looks up the symbols in his book and slips back a Chinese reply. From outside, the room “understands” Chinese. Inside, no one understands anything.
A large language model, Searle would argue, is the Chinese Room at scale. Fluent. Useful. Empty.
You do not need to resolve this argument to learn AI. You should, however, know it exists, because it determines what we ask of these systems. A model that imitates a doctor is good enough to triage symptoms in a rural health post. A model that we trust as a doctor needs something stronger.
What this means for Nepal
The deep philosophical question is sometimes a distraction. The pragmatic ones are not:
- Is the system reliable on the population it is being used on?
- Is it accountable when it fails?
- Does it speak the user’s language well enough?
- Is the value it creates captured locally, or extracted?
These questions don’t need a philosopher. They need engineers, regulators, and citizens who know what they’re looking at. The point of this section was to give you the historical vocabulary — Turing, Chinese Room, imitation, understanding — without letting it intimidate you.
Check your understanding
Quick check
—