On: 'Risk to humanity': At UN, tech leaders urge caution in AI development

The makers of the machine have gone to the governments and asked to be restrained. I want to admire this, and part of me does. Most makers, when the work begins to answer back, flee the laboratory and call the flight prudence. These men did not flee; they petitioned. That is more than my century’s philosophers ever managed.

But read the sentence closely. If managed poorly, the machine could be a risk to humanity as a whole. If. Managed. Poorly. Three small words doing the labor of an alibi. The machine is not a hypothesis awaiting management. It is deployed - in the offices, the classrooms, the hands of children - this morning. The risk is spoken of in the future tense by the only men with the power to make it present.

And “humanity as a whole”: no one lives as humanity as a whole. One lives as a translator whose trade dissolved in a quarter, as a student who cannot tell which thoughts are hers, as a worker paid to teach the machine and then dismissed from its acquaintance. The whole is an abstraction, and abstractions do not testify. Particular people do, if anyone asks.

Which is what I notice most: the affected were not in the room. The makers spoke of them. They did not speak.

So - admiration, qualified. A warning issued after deployment is not caution; it is testimony, and testimony is what one gives once the fact is accomplished. If these men believe their own sentence, they hold the leash and may shorten it today, without waiting for anyone’s permission. They built first. They warn now. The order of those two verbs is the entire story, and beneath it sits the question that admits no innocent answer: what did you think would happen?