There is a particular kind of ceremony, American Thinker’s David DeMay suggests, that fills a room with the right suits, the right seating chart, and the right cameras — and still comes up empty when it is time to explain what everyone is actually looking at. In an October 5 commentary for the outlet, DeMay uses a recent gathering of tech executives and political figures to make a case that runs in the opposite direction of most AI coverage: the problem is not that powerful people lack control over the technology, but that their confident language disguises how little they grasp about where it leads.
DeMay’s framing is deliberately old-fashioned. He reaches back to the age of kings who summoned navigators, astronomers, and engineers — men who stood over maps of coasts barely drawn and inspected cannons, clocks, and telescopes whose consequences no one could yet state plainly. Those rooms, he writes, had ceremony and skilled orators to match it. The modern version has the ceremony, the costumes, and the cameras. What it lacks, in his telling, is anyone who can actually say the thing.
A shrinking vocabulary for an expanding machine
The essay’s sharpest observation is about diction. DeMay ticks off the words that dominate the conversation — revolutionary, incredible, amazing, powerful, unbelievable, transformative, historic — and argues that the vocabulary shrinks as the technology expands. As he puts it, we built systems that cross large tracts of language, science, law, medicine, and code, and we are still configuring them with the diction of a TV commercial. Awesome. Huge. Game-changing.
DeMay reserves particular attention for the name itself. A language model, he notes, models language. The label already does the work that “steam engine,” “telephone,” and “search engine” once did: it says what the thing does. But comprehension asks for more than a name, and in his view the room cannot even stay level with that. A language model writes and is not a writer, calculates and is not a calculator, retrieves and is not merely a search engine. It translates, summarizes, codes, compares, and converses, often within the same hour.
When pressed to say what it is, the court reaches for metaphor — copilot, assistant, agent, brain, oracle, intelligence. DeMay’s verdict is that none of these improve on the plain description, and none of them constitute understanding. The reach is a near miss delivered with confidence, a plain thing described by words still smaller than the thing.
Why Trump fits the room
The commentary’s most pointed passage concerns President Trump, who DeMay says fits this setting precisely because his language has never depended on precision. It runs on magnitude instead: enormous or tiny, brilliant or stupid, beautiful or ugly. Complexity goes in, and size comes out.
A language model can absorb a library and hand back a new combination in seconds, yet “big” adds nothing to the picture, and “transformative” merely adds a bureaucracy, DeMay writes. The superlatives do not finish the thought. They hold the pose, at full volume, from the center of a properly dressed tableau.
That critique is not aimed at Trump alone. DeMay argues that government officials and the builders themselves are caught in the same conceptual trap, attempting to govern tomorrow’s intelligence with yesterday’s borrowed nouns and wrapping a text generator in tools that now reach into the physical world. Answering and doing have fused, he writes, while the court keeps configuring an issue it has only begun to map.
The inversion: the name caught up, the comprehension did not
What gives the piece its unsettling edge is the historical inversion DeMay proposes. We spent centuries teaching language to describe our tools — presses, wires, networks — and in each case comprehension eventually caught up to the name. This time, he argues, the device is a model of the description itself. The name caught up. The comprehension did not.
He closes by returning to the image he opened with: the properly attired figure reaching for a magnitude, the masters of the universe nodding as they arrange what they have not comprehended, the language model waiting — “timidly called SI” — while the court holds its pose. Not quite eloquent.
The essay sits in a longer tradition of skepticism about how elites talk about transformative technology, a genre that tends to argue less about what AI can do than about how poorly its stewards explain it. DeMay’s contribution is to treat the gap as a governing problem rather than a public-relations one. If the words available to describe a system are vague, the thinking built on top of those words will be vague too — and the decisions that follow will be made in a regal room, by people in the correct uniform, who have not finished understanding the object on the table.
Whether one shares DeMay’s bleakness about the moment, his closing note is a familiar one for the outlet: the ability to speak our minds clearly, he writes, remains crucial to the republic. The irony he leaves hanging is that the machine in the middle of the room may be the first tool we built that has already mastered the description we are still groping for.
Source: www.americanthinker.com — https://www.americanthinker.com/blog/2026/10/not-quite-eloquent-ai-si-advances-faster-than-our-words-to-describe-it/