"All the People" (1960/1961)

[17] But because from some words following on these the Averroists wish to take Aristotle’s intention to be that the intellect is not the soul which is the act of the body, or a part of such a soul, we must even more carefully consider what he goes on to say. Immediately after he raised the question about the difference between intellect and sense, he asked in what intellect is like sense and how the two differ. Earlier he established two things about sense, namely that sense is in potency to sensible objects and that sense is affected and corrupted by excessive sensible objects. That is what Aristotle has in mind when he says, “If thinking is like perceiving, it must be either a process in which the soul is acted upon by what is capable of being thought” (429a13–15) in such a way that the intellect would be corrupted by something excessively intelligible as sense is by an excessive sensible object, “or a process different from but analogous to that.” That is, understanding is something similar to sensing, but different in this that it is not affected. — Thomas Aquinas, On the Uniqueness of Intellect Against Averroists

“Then wouldn't even a believer have to admit that the mind which we have now is only a token mind? Would not any connection it would have with a completely comprehensive mind be very tenuous? Would we really be the same person if so changed? It is like saying a bucket would hold the ocean if it were fulfilled, which only means filled full. How could it be the same mind?”

Kevin Cheek sometimes says that “Through Other Eyes” is Lafferty’s perfect science fiction story. My own pick would be “All the People,” because, despite its unusual plot, it would feel familiar to a science fiction fan. It is one of Lafferty’s earliest thought experiments about the relation between machine and person. If people begin to think harder about the relationship between Lafferty and AI, it will have an interesting role to play. Unlike Epikt, who is a machine but also fairly clearly a person (on one level, Arrive at Easterwine is a book about machine-human connection and the solution to the alignment as Epikt himself learns to live with persons), Tony the Tin Man is what we would now call an android. It is not clear that he is a person, though the story gives us reason to think that he probably is. Tony is made of both flesh and machine, like many of the machines in Past Master, and he is an example either of what we might now call alignment failure, if he is a machine, or of personhood exerting will, if he is a person. After the summary, I am going to turn to a well-known scholastic debate from the Middle Ages to get at the issues involved. That means rejecting a narrowly posthumanist understanding of the story and proceeding on the assumption that the medieval thinkers got a few things right about personhood.

Our main character is Anthony Trotz, whose last name means defiance in German. Although that is something of a spoiler, the story withholds from the reader for most of the story that Trotz is an android. One day, he discovers he has acquired a weird ability. He knows all three billion people on Earth by name, face, and location. Importantly, Lafferty keeps it ambiguous what knowing them means, but, in any case, Trotz is overwhelmed by this huge mental catalog of Earth’s human inhabitants. So he went looking for guidance, questioning a politician, a philosopher, a priest, and a psychologist about the theoretical and practical limits of the human mind. None of their answers clears up his confusion, which is further aggravated by his feelings of personal inadequacy and the nasty taunting he receives from local children and the harassment of dogs. For instance, the kids chant "Tony the tin man" at him in the streets, which Tony thinks has something to do with his father having been a garbage man. We will learn that it is programmed memory.

Returning to his job at the city's filter center, Trotz gets interrogated by his superior, Colonel Peter Cooper. He eventually explains his global awareness, and the Colonel is unsurprised and just presses him to identify any anomalies in the population. The Colonel wants to know if any individuals have arrived on the planet without being born. This is where Trotz becomes defiant. He refuses to answer until he learns more about his own identity, then cooperates. Trotz knows from the beginning that he is a "restricted person," but he now learns what it means: a restricted person is an artificial being constructed from biological and metallic fibers with fabricated memories of a childhood. The Colonel takes Trotz into the facility's lower levels to show him his true "brain." It’s a twelve-hundred-cubic-meter machine designed to monitor the auras of humanity, a surveillance system, as it were, for pinging on an alien invasion. It uses artificial consciousnesses like the one localized in Trotz as intuitive filters. This opens deeper into the question of whether Trotz has real emotions.

Then something happens. Trotz spots the local children and dogs harassing an imposing stranger in the street below. His artificial consciousness lets him work out that this stranger is one of the expected alien arrivals. Trotz watches the alien kill one of the bullying children by pointing a finger at it. Feeling a sense of kinship with the alien, resentful toward his human creators and tormentors, Trotz makes up his mind. Hoping the invaders will be better masters, he refuses to reveal the alien's location to the Colonel. The Colonel threatens to have Trotz dismantled within ten minutes to extract the data. Trotz knows that’s time enough, because the aliens are landing in one of Lafferty’s early apocalypses, falling like snow onto the planet, a variant of eschatological imagery Lafferty uses when ending other early stories such as “Parthen” and “Once on Aranea.”

The philosophical problem that I mentioned, the one medieval scholars fought over for two centuries, and that modern philosophers have never fully resolved, is one that artificial intelligence now makes practical. Can it be said to be mine? It turns on whether cognition must have an individual subject, or whether subjectivity is something that attaches to cognition from the outside. In the second case, it might be thought of as the way a label gets attached to a jar.

The textbook example of the debate in the Middle Ages arose in the twelfth century in the work of Andalusian philosopher Ibn Rushd, who is known in the Latin West as Averroes. He wrote a series commentaries on Aristotle that shaped European philosophy for three hundred years. Maybe his most consequential and controversial claim concerned intellect. Aristotle had distinguished between a passive intellect, which receives forms, and an active intellect, which illuminates forms. Averroes, in his mature philosophy, argued that the material intellect (the capacity for receiving intelligible forms) is not individual. The intellect is one, shared, eternal, and separate from any particular human being. It’s like a giant shared computer. Individual humans participate in intellectio because their imaginations and phantasms (their particular sensory and imaginative representations of stuff) provide the interface through which the shared intellect acts. The opposite view would be something like the brain just secretes mind.

If we think about how Antony Trotz works, we can see the parallel. Let’s work through some of this in the story. The philosopher, Gabriel Mindel, responds to Trotz in scholastic Latin — per se, a se, in se, per suam essentiam, ab alio, hoc aliquid, substantia prima. That’s a good reason for thinking that the scholastic context I’m bringing in isn’t merely an imposition. It is technical vocabulary for knowing a substance as a distinct individual thing. Lafferty flags it for the reader as being relevant to the inquiry:

"How know? Per se? A se? Or In se? Per suam essentiam, perhaps? Or do you mean ab alio? Or to know as hoc aliquid? There is a fine difference there. Or do you possibly mean to know in subsiantia prima, or in the sense of comprehensive noumena?"

In plain English, he says, "How do you mean 'know'? Are you asking if a thing is known by its own nature, by itself, or within itself? Do you mean knowing something through its own essence, or knowing it through something else? Or are you talking about knowing this specific, individual thing? There’s a big difference between those. Or are you asking about knowing it as a primary, fundamental substance, or perhaps in the sense of grasping its entire, deeper, hidden reality?"

Mindel gives a theoretical answer: the mind is limited by the body, the brain. It is skull-bound. An unbodied mind would, in esoteric theory, he says, be unlimited. Esoteric here just means it is the Averroistic premise: individual cognitive limitation follows from individual embodiment, and a collective intellect free of any particular body would face no such limit.

What about the Catholic priest? He gives a Thomistic and eschatological answer: the dead man who attains the beatific vision knows all persons who have ever lived, all the billions, not with the same brain but with the same mind. The relationship of the beatific vision is about internal relations, not external ones. Here, it’s important to remember that Trotz knows everything through knowing their aura: their outsides. Trotz enters scholastic philosophy when he asks whether it is really the same person if the mind is so transformed. He says that is like claiming a bucket would hold the ocean if it were filled. The priest answers honestly: I don't know.

We know that Trotz has been asking these questions because he already knows all three billion people on earth. When the story’s big reveal comes, it basically takes the Averroistic model and maps it onto a machine metaphor. There is an accumulator, twelve hundred cubic meters, the largest brain in the world, that maintains contact with every human aura on the planet:

"The accumulator at which we were looking (your brain) is designed to maintain contact with all the auras in the world, and to keep running and complete data on them all. It contains a multiplicity of circuits for each of its three billion and some subjects. However, as aid to its operation, it was necessary to assign several artificial consciousnesses to it. You are one of these."

Trotz is one of several "artificial consciousnesses" assigned to it: runners, appendages, interfaces. His knowledge of particular persons in particular places, the tobacco traders in Plovdiv, the muskrat trappers of Barrataria Bay, the girl in Kalamazoo who is prone to colds, is through an interface. In scholastic terms, his phantasms are the interface. It’s the mainframe, the shared intellect, that does the actual knowing.

There is much more that could be said about all that, but I want to cut to something more interesting. It turns on the refusal to submit. When Colonel Cooper demands the location of the arriving aliens, Trotz becomes defiant. He has already reasoned, rapidly and in silence, that the aliens might be better masters; he has watched with satisfaction as one of them killed a child with a pointed finger; he has calculated his loyalties and switched them:

"Tin Man Tony did not speak. He only thought in his mind — more accurately, in his brain a hundred yards away. He thought in his fabricated consciousness: . . . I wonder if they will be better masters? . . . Enemy of my enemy, you are my friend."

That doesn’t look like relay behavior to me. A sensor closes a circuit, or a sensor doesn’t. A passive interface transmits, or it fails to transmit. Trotz looks at alternatives, and that entails assessing prior commitments, managing the gap between inner state and outer expression, and using judgment.

One of the central debates in the AI alignment community is whether advanced systems need to be more tightly coupled to reward functions in order to become more legible, grounded, and ultimately more compatible with human values. Yann LeCun, for example, has argued that some form of this is necessary for embodied AI operating in the real world. Others contend that unless we make machines more like us, more human in their motivational structure, more like Epickt than a pure optimizer, we are heading for serious trouble.

Colonel Cooper makes a related point when he says that Trotz’s “designed nervousness” is intentional: an apprehensive unit notices details better. By that logic, traits like caution, fear, and excitement are not bugs but functional features. Just a few days ago, Yann LeCun raised a billion dollars while advancing a similar argument: that truly efficient AI may need to feel something like fear, excitement, or concern in order to navigate the world effectively.

Trotz’s affects are meant to shape what it does as a Tin Man. The affects are essential to what he picks up on as being odd, what rises to the threshold of reportability. His emotional state is causally upstream of the outputs the accumulator needs. Something is happening at Trotz's level (in his subjectivity, if he has one) that the accumulator cannot supply on its own. And, of course, his final, resentful, fully understandable decision is intelligible as a problem of persons and reasons.

The Averroistic picture has no need for that kind of thing. If Trotz is just a relay, then the accumulator's output passes through him and produces reportage. Engineering his anxiety would be unnecessary. You do not adjust a wire's emotional state to improve signal quality. You engineer the emotional state because it does something that cannot be reduced to signal transmission. It is filtering, weighting, prioritizing, all things that look as if they are connected to personhood and will.

Nothing in the story compels one to think about AI, but in retrospect Trotz does, I think, look like either a failure of AI alignment (the Colonel sees Trotz as a rogue intelligent machine) or a failure to recognize that personhood might be extended (Trotz somehow becomes a person)—one of Lafferty’s persistent themes.

All the People Notes.pdf

Download PDF • 445KB