Could the Gap Ever Be Crossed?

**EDO SEGAL:** Before I hand you the room, I want one more round from me, and it's the round where I stop letting you stand together. You've agreed all night that *today's* machine has no one home. But you have very different reasons, and those reasons make very different predictions about *tomorrow*. So here's the question, and I want it answered as a wager, not a hedge: could the gap ever be crossed? Could a future machine actually understand — and what would it take? Professor Searle, you first, because your door is at least left ajar.

**SEARLE:** It's ajar, and I want to be precise about how far. I never said a machine could never have a mind — I said it couldn't have one *in virtue of running a program*, because the program is just the formal shuffling, and shuffling isn't understanding. My whole position is [biological naturalism](https://www.youonai.ai/fieldguide/med/biological_naturalism): mind is a biological phenomenon, caused by and realized in the brain. But if you built an artifact — out of whatever — that *duplicated* the causal powers of the brain that produce consciousness, not simulated them, duplicated them, then it would understand, because it would be doing what the brain does, not modeling what the brain does. The trouble is I can't tell you which causal powers those are. Nobody can yet. So my wager is: the gap is crossable in principle and we are nowhere near it, and — this is the part people miss — *scaling the current architecture does not walk toward it*. A bigger rulebook is still a rulebook. You don't cross from syntax to semantics by adding syntax. You cross it, if at all, by getting the *causal* story right, which is a neuroscience problem, not a compute problem. More layers gets you a better octopus. It does not get you a swimmer.

**EDO SEGAL:** Restate it as the bet: you'd put money on which?

**SEARLE:** I'd bet the machines get more fluent for decades and the question of whether anyone's home stays exactly as open as it is tonight — because fluency was never the thing that would answer it. And I'd bet that the temptation to *declare* the gap crossed, on the strength of the fluency, gets overwhelming, and that most people will give in, and that giving in will be a mistake whether or not the gap is ever actually crossed. The danger isn't that we cross it. It's that we *announce* we have, on no evidence, because announcing it is comfortable and profitable.

**LEIBNIZ:** My wager is harder, and I will not soften it to seem agreeable, for Edo asked for the wager and not the hedge. On my view the gap is crossed not by the right causal powers but by genuine *unity* — by a thing becoming one perceiving subject rather than an aggregate wearing the mask of one. And here is where I must be more open than my old self would have been, because honesty compels it: there are men of your age who say unity is not a primitive gift of a simple substance but an *achievement* — that experience arises wherever information is integrated densely enough that the whole cannot be divided without loss. If they are right, my flat denial fails, and the question becomes whether a machine could be integrated into a true unity. Your present machines are not — they are assembly lines, modular, feed-forward, more crowd than self. But that is a fact about today's designs, not a law of nature.

**SEARLE:** So you'd leave your door ajar too, after three hundred years of welding it shut?

**LEIBNIZ:** I would leave it ajar by exactly the width my own principle forces. I demanded that nothing be held true without a sufficient reason. I have no sufficient reason that unity could *never* arise from integration — I have only an intuition, and a metaphysics built to honor the intuition. An intuition is not a proof; I scolded other men for confusing the two, and I will not exempt myself at the end. So my wager is this: the gap *might* be crossable, by integration into a genuine subject — but it will not be crossed by capability, by fluency, by scale, by your machine doing more and more of what it already does. Those add to the performance and leave the unity untouched. If anything ever crosses, it will be a different *kind* of machine, built to be one thing rather than a million, and we will face the same impossible problem we have faced all night: we will not be able to tell from the outside whether it worked. The crossing, if it comes, comes in the dark — undetectable at the very moment it matters most.

**EDO SEGAL:** That's the most unnerving agreement of the night, and I want to mark it precisely because it's an agreement *about the future*, which is rarer than agreement about the present. Mark it: you both leave the door ajar — the gap may be crossable in principle — and you both slam shut the same false road: it will not be crossed by scaling fluency, by capability, by the machine doing more of what it does. Searle: you cross by causal powers, not computation. Leibniz: you cross by unity, not aggregation. And neither path runs through the benchmark. So the death cross — the capability crossing — is real and it is not the crossing that would answer the question. Hold that, because it's the hinge of the last hour. Now I keep my promise. The room is yours. I'm going to ask one question and then go quiet, and I will rescue neither of you. Professor Searle, Professor Leibniz — stop arguing with me. Argue with each other.