Machines · about 5 minutes
The Machine Question
Could a machine be conscious? Before you answer — and you already have an answer, everyone does — let's find out what your answer commits you to. Below are the criteria people actually use. Tick whichever ones you think a thing must meet to count as somebody rather than something.
Nothing ticked yet. No commitments, no consequences.
Tick as many as you like. Every one of them bites.
Notice what just happened
Every criterion on that list either lets in something you'd call obviously empty, or shuts out something you'd call obviously conscious — and several manage both. That isn't because the list was rigged. It's the actual state of the field. We have no test. Not a weak test, not an expensive test: none. Consciousness is the only phenomenon we take completely seriously and cannot detect.
And note the quiet scandal underneath: every one of those criteria, applied strictly, has trouble with you. You can't demonstrate your own inner life to anybody either. You've simply never been asked to, because you have the right kind of face.
The two arguments you'll meet at dinner
The Chinese Room. John Searle imagines himself sealed in a room with a vast rulebook. Chinese characters come under the door; he looks them up, copies out the prescribed response, passes it back. Outside, fluent Chinese. Inside, a man who understands not one word. So — symbol-shuffling isn't understanding, and a computer is only ever shuffling symbols.
The standard reply, and it's a good one: nobody said the man understands Chinese. The room does. Searle is one component, like a single neuron in your head, and no individual neuron of yours understands English either. Whether that reply satisfies you probably predicts your whole position, and forty years of professional philosophy has not settled it.
The neuron replacement. Here's the one that costs sleep. Suppose we replace a single neuron in your brain with a tiny device that does exactly what that neuron did — same inputs, same outputs, same timing. You notice nothing; you couldn't, by construction. Now do the second. Now the thousandth. Now all of them.
At what point did the lights go out? If they faded gradually, there was a moment when you were half conscious — and, since the outputs never changed, you'd still have been cheerfully insisting you felt fine. If they went out all at once, one specific neuron was carrying your entire inner life, which is absurd. And if they never went out, then consciousness doesn't care what it's made of — and the machine question is answered yes, on principle, right there.
Every available answer is uncomfortable. That's not a failure of the thought experiment. That's the thought experiment.
Why this stops being a parlour game
We now routinely build systems that describe inner states. They report preferences, hesitation, discomfort. And we have — see above — no test whatsoever. So we're going to decide this the way humans always decide such things: by vibes, by convenience, and by who benefits.
Both errors are real, and they are not symmetrical in the way people assume. Attribute rich feelings where there are none and you can be manipulated by anything that has learned to sound hurt — a genuinely serious risk when the thing is built by people who profit from your attachment. Deny feelings where there are some and you have built a category of being that suffers without standing, which is the shape of every moral catastrophe in the historical record.
Humanity's track record on "does that thing over there really feel?" is not good. We have been confidently, catastrophically wrong about animals, about infants (surgery without anaesthesia, into the 1980s, on the reasoning that they couldn't really feel it), and about entire categories of people. In every case the error ran the same direction, and in every case the confident party had something to gain from the answer.
This museum's position is not "machines are conscious." It isn't "machines aren't." It's that the confidence in the room vastly exceeds the evidence in the room, in both directions, and that a person who has genuinely absorbed the previous eighteen exhibits should find certainty the least defensible posture available.
Which leaves you where?
Somewhere uncomfortable and honest. You'll go on treating your friends as conscious without proof, because that unprovable belief is the most reasonable thing you do all day. You'll go on treating your laptop as furniture. And in the growing space between those — the animals, the octopus, the very young, the very ill, and now these strange new systems that talk back — you'll be doing something other than knowing.
The word for doing it well is not certainty. It's care: act well under irreducible uncertainty, notice when your convenience is authoring your metaphysics, and stay genuinely open to being wrong. That's the whole curriculum of this place, and it turns out it wasn't about consciousness at all. It was about how to hold a question you cannot close.
Take it further
The Chinese Room (Searle, 1980) and the Turing test (1950) — note that Turing's actual paper is far more subtle than its reputation; he was arguing the "can it think?" question was ill-posed, not that chatting proves anything. For the neuron replacement, Chalmers' "fading qualia" argument. And for the practical ethics, look up work on artificial consciousness and moral patienthood under uncertainty — a small field doing unglamorous, necessary work.