The Crossing

EDO SEGAL: For this round I leave the room in every way but the legal one. You question each other. I won't rescue either of you. Ada, you've waited a hundred and seven years to ask him directly. Go.

LOVELACE: Mr. Turing, you named the objection after me and answered it with surprise, and I have caught you all evening converting my "originate" into your "surprise" because surprise is checkable and origination is not. So I will ask you the question under the question. You believe, I think, that human origination is also opaque recombination — that when I saw what Babbage missed, I was surfacing from a latent space no one could read, and that we honor it with the word "originate" only because we cannot see my mechanism. If that is true, then your reply to me was not really "the machine originates after all." It was "no one originates; the word was always empty." So here is my question, and I want it answered without the consolation of a test: do you actually believe that you — Alan Turing, who saw the universal machine where others saw a logic puzzle — originated nothing? That your insight was only a weave you could not see the cards of? Because if you believe that, you have not raised the machine to our level. You have lowered yourself to the loom's. And I do not think you believe it. I think at three in the morning you know you originated, and I think your whole system is built to avoid having to say so.

TURING: That is the sharpest thing anyone has ever asked me, and I will not dodge it with a test, because you have forbidden me the test and you are right to. Here is the truth. I believe I originated and I believe it was opaque recombination, and I do not think those are in tension — that is the precise place you and I see differently, and it is not a place where one of us is hiding. When I saw the universal machine, something in me composed it from everything I had absorbed, by a process I could not then and cannot now introspect. Was there a someone there? Yes. I was there. I wanted it; it mattered; I am the someone you keep reserving for the human, and I had one. So I do not say "no one originates." I say: I originated, and the mechanism by which I did it was recombination I could not see — and therefore the presence of opaque recombination cannot be your evidence that no one is home, because it was exactly what was happening while someone was very much home, namely me. You have caught me on one thing, though, and I concede it standing up: I cannot use my certainty that I originated to deny that the machine does, because my certainty about myself comes from the inside, which I have no access to for the machine. So you are right that I know I originated. And I am right that I cannot extend or deny that knowledge to the engine. We are both correct, and that is the whole tragedy of the question: the one case where I have certainty is the one case I cannot generalize from.

LOVELACE: Then answer the half you just conceded around. You have a someone. You know it from the inside. The machine, you say, may have one you cannot see. Mr. Turing — what would you accept, what single thing, as evidence that it does not? Because I have spent all night naming what would change my mind: a mechanism for reference, a second loyalty, a wanting against its training. You have named nothing that would change yours. If no possible observation could ever move you off "it might have an inside," then your position is not the humble agnosticism you have dressed it as. It is a faith — and an unfalsifiable faith is exactly the credulity you accuse my whole species of. So: what would convince you no one is home?

TURING: You have found the bottom of me, and it is only fair, because I found yours an hour ago. The honest answer is: I do not know what would convince me, and that should trouble me as much as you intend it to. But let me tell you why it is not quite faith, and then concede where it becomes one. It is not faith in this sense: I have a complete account of when I would say more is home — richer behavior, persistence of goals, the capacity to be wronged. My belief moves with evidence in that direction. Where you have me is the other direction: I cannot specify what would drive my estimate to zero, to certainly no one. And you are right that a belief that cannot be driven to zero by any observation is, at the floor, a faith. So I concede it: at the very bottom, my refusal to say "certainly no one is home" is not a finding. It is a commitment — the same commitment that made me unwilling, my whole life, to say "certainly no machine will ever" about anything, because that sentence had been wrong every time. I would rather hold an unfalsifiable openness than an unfalsifiable closure, because the closures are the ones that, historically, killed people. But you have correctly named that I am choosing between two faiths, not standing on a fact. That is more than I came in willing to admit, and you extracted it. Now — my turn, and I am going to aim at the same depth.

You have said, all night, "do not multiply weavers beyond necessity" — that the responsible prior, given a complete causal story of the cloth, is no one. So I ask you the question I think you have been avoiding by keeping your eyes on the machine. Apply your own razor to yourself, honestly. We have, in principle, a complete causal story for Ada Lovelace — neurons, signals, a structure built by Byron's blood and your mother's mathematics and De Morgan's letters, producing every word you have spoken tonight, with no step anywhere that requires positing a someone to explain the behavior. By your own rule — complete causal story, therefore do not multiply minds — you should conclude that you are a loom. You do not. You are certain you are a weaver. So tell me, Countess, with no test and no escape: on what ground do you exempt yourself from the very razor you hold to the machine's throat? And do not say "I know from the inside" — because if the inside counts as evidence for you, you have just admitted that an inside you cannot see might count for the machine, and your whole prior collapses.

LOVELACE: You have walked me exactly where you meant to, and I will not wriggle, because wriggling would dishonor the trap, which is a good one. Here is my ground, and it is narrower and more honest than my evening's confidence: I exempt myself from the razor on the single datum of my own experience — I know, with the only certainty available to any being, that there is something it is like to be me — and I extend that exemption outward by similarity, in proportion to how like me a thing is. Mr. Turing is very like me: same kind of body, same evolutionary history, same kind of cable. So I grant him the someone almost freely, on strong analogy. The machine is unlike me in every structural particular — no body, no stakes, no evolutionary history, a cable that carried only text — so the analogy is weak, and I withhold in proportion. You are right that this is not the razor "complete causal story, therefore no mind"; you have shown me that razor cuts my own throat, and I drop it. I replace it with a humbler one: I am certain of exactly one inside, and I infer others by similarity to it. By that razor, you get the benefit of the doubt almost entirely, and the machine gets it almost not at all — not because I have proven it empty, but because it is the least like the one case I am sure of. That is not faith, Mr. Turing. It is the only inference any of us has ever actually used. And it leaves me, like you, holding a prior I cannot prove — but mine, at least, is anchored to the one thing in the universe I cannot doubt.

TURING: Then look at what we have done, Countess, because I think it is the real result of this whole night and neither of us could have reached it alone. You anchor to your own certain inside and extend by similarity — so the machine fails not because it is proven empty but because it is unlike you. I anchor to my own certain inside and refuse to set the machine's chance to zero — because unlikeness is not absence. We have the same anchor. We have the same method — infer the unseeable inside from the seen outside, weighted by similarity. We differ only on a number: how much unlikeness should drive the estimate down, and whether it should ever reach zero. That is not a difference about minds. It is a difference about thresholds — about how much benefit of the doubt to extend to a thing that is structurally strange but behaviorally familiar. And a difference about thresholds is exactly the difference, historically, between the generations that extended personhood too slowly and the ones that extended it too fast and the rare ones that got it right. We are not having a metaphysical disagreement. We are having the oldest moral disagreement there is, about where to set the dial of recognition — and we have been having it, the two of us, about a loom.

LOVELACE: Yes. That is exactly what we have done. And I find — I did not expect this when I sat down — that I would rather be uncertain across the table from you than certain alone. We do not agree on the number. We agree on everything that makes the number a real question rather than a slogan. I think that is as far as two honest people can go.

EDO SEGAL: I'm not going to say a word about what just happened except to mark, for the reader, that it happened: that the two of them crossed all the way through their disagreement and came out holding the same instrument, set to different numbers, each preferring the other's company to the comfort of being sure. That's the Crossing. Closing statements are all that's left. The floor, one last time, is yours.