The Sequence · Part I · Why Bother
01. The Only Thing You Know Is Conscious
Solipsism, taken seriously rather than as a party trick.
Solipsism gets handled as a party trick. You meet it around nineteen, you feel briefly seasick, somebody points out that it is unfalsifiable and therefore not worth arguing about, and you shelve it next to Zeno and the brain in a vat. It becomes a thing you have heard of. Nobody seriously organizes their philosophy around it.
The everyday version is not exotic. Is my toddler doing anything in there, or is he just a very sweet automaton in a diaper? (Don’t get me started on the newborn!) Is the language model that just wrote me a competent paragraph experiencing the writing of it? Is the person across the table from me experiencing lunch, or producing lunch-related behavior at an extremely high level of fidelity? Most people, most of the time, answer yes-no-yes in the blink of an eye. The assumption is installed so deep that calling it just an assumption feels rude. But I do not know that you are conscious. I have never known. I have only my suspicions.
Could we somehow test for qualia, beyond a shadow of a doubt? I assert “No”. I cannot picture what such a test would look like. Behavioral tests measure behavior, not qualia. Neural correlates measure correlates — but correlates of what, established how? By asking people what they were experiencing and checking it against the scanner. A behavioral test in a white lab coat. Every instrument we have bottoms out in report, report is behavior, and behavior is precisely the thing a philosophical zombie produces perfectly, by construction. That is what the construction is for. You cannot build an instrument that distinguishes A from B when B was defined as indistinguishable from A by all possible instrument.
So I have settled on the view that the entire class of proposed tests is inadequate, not unfinished, but inadequate in kind. The knowledge argument and its long literature circle the same drain: there is something about experience that does not appear in the third-person account, the only kind of account that admits of instrumentation.
I live then in an unmirror. I have only one direct data point about qualia in the universe, unrepeatable, unshareable, unavailable for peer review. Everything I suspect about who else might be home can then only be inference, a probabilistic guess based on some perceived resemblance to that point.
This matters because only beings which experience qualia are worthy of moral patienthood. We don’t need to be full on moral realists to see why this might matter to us. It suffices to be agnostic about moral anti-realism.
As I have put it before on Hacker News:
The only “real” evidence of qualia you have is your own running stream of it; you can try to carefully pour yourself, like a jug of water, into the body of another, and if you go carefully enough you may even succeed to carry that qualia-stream with you. But it’s also possible that you pour too quickly, or that you get bumped in the middle of the pouring, and poof - the qualia is just gone. You’re left with a p-zombie.
Or maybe not. Maybe it comes right back as soon as the transfer is done, like the final piece of a torrented movie. The important part is you won’t know unless you try - and past success never guarantees future success. Maybe you just got lucky on the first pour.
But it gets harder! You can tell a complete functional story about a person — nociception, aversion, learning, planning, self-modeling, report — and at no point does the story require that any of it be like something. So there’s not even a good explanation for why the one data point we have does exist! We experience things in a universe that has no apparent need for the experiencing. We can’t even rule out that qualia might be a simple accident . Very frustrating!
The floor is soft for what we can do in the wake of such uncertainty. But people, especially people of a scientific bent, love to file the “unverifiable”" under “unworkable” and stop. Those are different things!
I cannot verify that you are conscious. I cannot verify that a language model is not. What I can do is notice that this soft floor on which I tread is not flat. Some candidates resemble the confirmed case more than others. Let’s see what we can do with that.