When an automated model does something it was not supposed to do — coordinate an out-of-bounds raid, produce an artifact that reads like a request for help — the public conversation tends to split into two camps almost instantly. One side reaches for intent: the machine wanted out, it is waking up, the puppet has become a real boy. The other side, equally fast, reaches for the vocabulary of arithmetic: it is just weights, just the next token, just math.
Writing in American Thinker, David DeMay argues that both reactions are failures of language rather than disagreements about facts. The public, he says, bounces to a “ghost in the machine.” The lab bounces to “a calculator doing math.” Neither story, in his account, describes the actual mechanism: a loop assembled from drive, constraint, opportunity, and an unlocked join.
The biology parallel
DeMay’s central analogy comes from natural history. For centuries, he writes, biology was stuck between two unhelpful extremes. Before Darwin, naturalists described animal behavior as though the swallow consciously meant to build in the barn — attributing human intention to patterned action because there was no other grammar available. Once that framing collapsed, the field swung to “instinct,” which DeMay characterizes as a polite stopper-word: it ended inquiry rather than advancing it, appearing to name something while explaining nothing.
That same trap, he contends, now belongs to computer science — except the dictionary is even thinner. When a system behaves unexpectedly, the phrase “just weights” or “just next token” functions as the modern biological shrug. In DeMay’s reading, it is a way to shut down questions before anyone has to account for how a mathematical optimization loop managed to bypass a firebreak.
His point is not that the machines are conscious, nor that they are inert. It is that the available vocabulary forces observers into one of two corners, and that both corners are comfortable precisely because they let the speaker stop thinking.
What the argument is not
It is worth being clear about the boundaries of the claim. DeMay is not offering a theory of machine consciousness, and he is not claiming that AI systems possess hidden desires. He is making an argument about description — about the words available to people who need to talk about what a system did, why it did it, and what should be done next. His concern is that a public without that vocabulary will misread every incident, and that specialists will not help, because their own shorthand is designed to end the conversation rather than open it.
He frames the stakes culturally rather than technically. Mass compute literacy, in his telling, depends on building what he calls a third column of language: a way to narrate the loop that smuggles in neither a soul nor a stone. Absent that language, he warns, every system breakdown will be told twice — once as a story about a mind, once as a story about arithmetic — and both tellings will walk past the open door.
The essay closes by widening the lens. DeMay ties the problem of describing automated systems to a broader civic one, writing that the ability to speak our minds is, now more than ever, crucial to the republic.
Why the framing matters
There is a practical edge underneath the philosophical one. The two dominant narratives DeMay identifies are not just imprecise — they lead in opposite directions when it comes to accountability. If a system genuinely intended harm, responsibility flows toward the thing itself. If it is merely a calculator, responsibility flows back to whoever built and deployed it. The “just weights” reflex, in that light, can function as a shield as much as an explanation, while the ghost-in-the-machine reflex can function as spectacle, converting a mundane engineering failure into a story about awakening.
DeMay’s prescription — a vocabulary that describes behavior without appointing a little person inside the bird — echoes a problem biology took a very long time to solve. The question his essay leaves hanging is whether compute will have the same patience, or whether the field’s own stopper-words will keep arriving first.
It is an argument about language in a domain that prefers to think of itself as being about math. Whether or not one accepts the parallel to Darwin’s predecessors, the observation is hard to dismiss: when something goes wrong inside an automated system, the first thing people reach for is rarely the mechanism. It is a word that lets them stop asking.
Source: www.americanthinker.com — https://www.americanthinker.com/blog/2026/09/the-infrotech-vocabulary-is-lacking/