Mind and body · Quiz question 29/35

The AI That Says It's Sad

The Turing Test and the Chinese Room

An AI chats like a human and says it feels sad. Is it actually sad? The Turing test, John Searle's Chinese Room, and today's AI consciousness debate.

As it appears in the quiz

An AI chats so naturally you can't tell it from a person — and it tells you it feels sad right now.

That AI might still feel nothing at all.

Who came up with it

In "Computing Machinery and Intelligence" (1950), British mathematician Alan Turing proposed replacing "can machines think?" with an "imitation game": can it converse indistinguishably from a person? In 1980 American philosopher John Searle fired back with the "Chinese Room": someone who knows no Chinese follows a rulebook and produces perfect Chinese replies without understanding Chinese — so running a program alone produces neither understanding nor consciousness.

Original papers & key readings

What it's really asking

If you agreed, you think behavior can be identical while the inner experience is absent. That's Searle's Chinese Room intuition — and more broadly the "problem of other minds": behavior alone can't tell you whether something has a mind.

The most famous reply to the Chinese Room is the "systems reply": the person doesn't understand Chinese, but the whole system — person, rulebook, paper — might. With large language models, this debate went from hypothetical to real. In a 2022 talk, David Chalmers judged it unlikely that current language models are conscious, but said systems worth taking seriously as conscious could arrive within a decade. Many researchers agree that an AI saying "I'm sad" isn't strong evidence by itself, since it may just be reproducing human expressions from its training data.

Where philosophers stand

Chinese Room / biological naturalism (agrees)

Symbol manipulation doesn't produce understanding or feelings. Behavior isn't evidence of a mind.

Key names: John Searle, Thomas Nagel (on subjectivity)

Functionalism / behavioral criteria (leans no)

If it functions indistinguishably from a person, there's little ground to deny it a mind.

Key names: Turing, Daniel Dennett, Paul Churchland

Open question

Today's AI is unlikely to be conscious, but different architectures could be in principle. Look at the internals, not the words.

Key names: David Chalmers

In the 2020 PhilPapers Survey, about 67% of philosophers said the Chinese Room system doesn't understand Chinese, and about 39% said future AI systems could be conscious.

Philosophers to check out

Sources