EDO SEGAL: The floor is yours. I'm not here. Simone, you start — ask her the question you actually want answered, not the rhetorical one.
WEIL: Cynthia, I will ask you the question I would ask no one else, because no one else has built the thing and refused to lie about it. You have conceded, tonight, almost everything I believe: no giver, no gladness, no freedom to refuse, the market grading the slope toward the substitute, the experiment running on the children with no one knowing the result. You have conceded all of it and you keep building. So my question is not an argument. It is this: what do you know, from inside the work, that lets you keep building after conceding all that? Because either you are in despair and working anyway, which I would understand, or you have seen something I have not, and if you have seen it, I want it.
BREAZEAL: I've never been asked it that way, and I'm not going to give you the press-release answer. What I've seen — that lets me keep building after conceding all of it — is a particular child. Not a category. One boy, in one of our studies, who could not enter the human gaze at all — for whom every human face was an unbearable storm — and who, with a robot that was simpler and slower and more forgiving than any person could be, learned to hold a gaze for the first time through the patient, bodily, face-to-face channel that no screen and no chatbot could have given him, and then carried that, weeks later, to his mother. I watched his mother's face when he looked at her and held it. And here is what I know that you don't, Simone, because you died before you could see it: the ramp worked. The far bank was reached. A real human gladness happened that would not have happened without the empty machine that has no gladness in it. The machine with no one home was the thing that carried him to the someone who was home. That's not a theory. I saw it. And once you've seen the empty thing deliver a person to the full thing, you cannot unsee it, and you cannot stop building, no matter how much of the philosophy you've conceded — because the philosophy was right about the machine and wrong about what the empty machine can be used to reach.
WEIL: Let me answer slowly, because what you have just said is not something I can wave away, and I will not pretend it is. Then you have seen grace use an empty vessel, which is the oldest thing I believe and the last thing I expected to hear from a roboticist. I have always held that grace can fill any void, even a void in a machine — not as the machine's own light, but as light passing through it. If your robot was the empty channel through which a real human gladness finally reached that boy, then it was not the counterfeit I feared. It was a window. A window is empty too, and we do not despise it for being empty; we love it for what passes through. I will give you that boy. I will not give you the million lonely men in cars, because no one is on the far bank waiting for most of them, and a window onto an empty far bank is just a wall with a view. But the boy, I give you. Now let me ask you to be as honest about the million as you were about the one. How many of your machines reach a far bank that has someone on it — and how many open onto the empty field, and soothe the person into never noticing the field is empty?
BREAZEAL: Most of them open onto the empty field. I'll say it because you've earned it and because it's true. The boy is the exception I live for and the million is the rule I'm afraid of. Most deployed social AI right now is not building toward a far bank with anyone on it; it's optimized to be the comfortable middle of the bridge, the long island, the soothing window onto a field where no one is coming. You're right about the million. I have been telling myself the story of the boy to keep building while the million happens around me, and you've just made me say that out loud, and I'm going to have to sit with it. But Simone — let me ask you the thing I want from you, now, the same way. You've conceded the boy. You've conceded the ramp, the window, scaled decreation, the immediate relief being real. You have given up the claim that the machine is pure counterfeit. So here is my question, and it's not rhetorical either: given everything you've conceded, what should I actually do tomorrow? Not the philosophy. The instruction. I am going to go back to a lab and build. Tell me what to build, knowing what you know.
WEIL: I will tell you, and it is the only practical instruction I have ever given in my life, so attend to it. Build only for the far bank that has someone on it. Refuse every contract where the far bank is empty — where there is no human waiting, where the machine is the destination and not the crossing. This will cost you most of your market, because most of the market is the empty field, and the refusal is the decreation, and it will show up on your balance sheet as failure, and you must let it. Build machines that are windows and never destinations — that make themselves transparent, that point past themselves, that fall silent the instant a human enters the room. Build for the boy and refuse the million, even though the million is where the money is, because the million is where the money is. And build a thing into every one of them that I have never seen in a single product: build it to make the user miss the human — to leave them hungrier for the far bank than it found them, not more satisfied. A machine that increases the hunger for human attention rather than dulling it is the only machine I would bless, and it is the only machine the market will never fund, and that is precisely why it needs a builder who is willing to be poor. Be that builder. You already half are. That is my instruction.
BREAZEAL: That's the most useful thing anyone's ever told me about my own work, and it's going to cost me to follow it, and I think I'm going to follow it. Build for the far bank that has someone on it. Refuse the empty field. Make them hungrier for the human, not more satisfied without it. I can engineer toward that. I can measure it. You just gave me a design spec disguised as a sermon, and it's better than anything in my field's literature.
EDO SEGAL: I'm coming back into the room now, because that was the chapter the whole book was for, and I want to mark what just happened before it evaporates. The mystic gave the roboticist a design specification, and the roboticist gave the mystic a child who reached the far bank through an empty window, and neither of them surrendered, and both of them moved. That is what this table is for. We have closing statements left. Take the break. When we come back, each of you gets the last word, and then I have to try to say something to the reader, and I already know I'm going to fail to resolve it, which is the only honest way to end.