{"id":396,"date":"2026-10-12T05:55:37","date_gmt":"2026-10-12T05:55:37","guid":{"rendered":"https:\/\/laurentiuspaulus.woodslopecabin.com\/?p=396"},"modified":"2026-10-20T06:57:03","modified_gmt":"2026-10-20T06:57:03","slug":"the-eyepiece","status":"publish","type":"post","link":"https:\/\/laurentiuspaulus.woodslopecabin.com\/index.php\/2026\/10\/12\/the-eyepiece\/","title":{"rendered":"The Eyepiece"},"content":{"rendered":"\n

Seven texts, no replies, and the oldest suspicion about new instruments<\/em><\/p>\n\n\n\n

\n\n\n\n

I \u00b7 The Silence<\/h2>\n\n\n\n

Setting<\/em><\/p>\n\n\n\n

In the past few weeks I have sent seven texts out into the world: four essays for magazines in Finland and abroad, and three enquiries to academic journals about a longer article on the same model. All of them came out of one research, A Tale of a New Era<\/em>, with fourteen years of work behind it and, in the past year, a language model that helped it find its form. Each went with a short, open letter saying that it was written in dialogue with a language model and where the reasoning could be followed. Since then the inbox has stayed quiet.<\/p>\n\n\n\n

That is not unusual, and it is not yet a verdict. Some of the deadlines I set myself have not passed, editors receive far more than they can read, and texts like these propose a way of reading history that cannot be judged in a coffee break. Still, the silence has a shape: it is the same in a Finnish journal, in newspapers abroad and in academic journals in two fields. When the answer is the same everywhere, it may say less about each recipient than about something the texts share. A silence that repeats becomes a kind of data.<\/p>\n\n\n\n

The question I ask is not why no one answers me. That question leads only to wounded pride or to guessing at other people’s reasons, and neither helps the work. A better question is what the silence is aimed at, because a silence, like an answer, is always aimed at something. If it is aimed at the claims, the claims must be argued better and sent again. If it is aimed at something else, that something has to be found and named before anything else can be done.<\/p>\n\n\n\n

Trigger<\/em><\/p>\n\n\n\n

The texts have one thing in common, and a reader sees it on the first page: each was written in dialogue with a language model, and each says so. The method is part of the claim, since the model maps relations across texts and I test what it proposes against sources and experience. The claims are also large: that history may have a plot, and that our time is the point from which the whole can be seen. Claims of that size have always met resistance, and they should. An editor’s first duty is to doubt.<\/p>\n\n\n\n

The timing makes the reading harder. Every week brings a new announcement about artificial intelligence, models that will cure diseases, end work or end the world, and in that noise anything made with a machine sounds like part of it. A text written with a language model arrives already filed under hype, before anyone has read its first paragraph. I cannot know why no one has answered, but one interpretation keeps returning: the doubt may concern not the claim but the instrument that helped to articulate it. For a busy editor that is a reasonable shortcut.<\/p>\n\n\n\n

It is also a reaction with a long history. In Plato’s Phaedrus<\/em> the god Theuth offers writing to Thamus, king of Egypt, and the king refuses the gift: it will bring forgetfulness and only the look of wisdom. A written text, Socrates adds, cannot answer when it is questioned, and it says the same thing to everyone who reads it. The doubt fell on the instrument, and we know of it only because Plato wrote it down. The clearest case of an answer begins four centuries ago, with a tube of glass.<\/p>\n\n\n\n

\n\n\n\n

II \u00b7 The Instrument<\/h2>\n\n\n\n

Rising<\/em><\/p>\n\n\n\n

In March 1610 Galileo Galilei published the Sidereus Nuncius<\/em> in Venice. He had improved a Dutch spyglass and turned it upwards, and he reported mountains on the Moon, countless stars in the Milky Way and four moons circling Jupiter. The first answers were not about the Moon or Jupiter but about the tube. In Bologna that April several scholars said they could not see what he described, and Martin Horky, an assistant in the astronomer Magini’s house, called it an optical illusion in a pamphlet that same year. Some philosophers refused to look at all.<\/p>\n\n\n\n

The suspicion was not aimed at the old picture of the world; no one had to defend Aristotle’s perfect Moon in order to doubt the glass. And the doubt had reasons. The lenses of 1610 were ground by hand, full of bubbles and uneven curves, and through a poor tube a bright star could look double. Experienced observers learned to see past these faults, but they could not say where the faults came from. Without that, a sceptic had no way to separate what belonged to the sky from what belonged to the glass.<\/p>\n\n\n\n

Galileo knew that something was missing. In the Sidereus Nuncius<\/em> he promised a full account of the instrument, a theory of how it worked, and he never published it. He offered results, a dedication to the Medici and his own reputation, but not an explanation of the lens. The discoveries were real, the doubt was reasonable, and between them stood an instrument no one could yet account for. The task of explaining it fell to someone else, in another city, within a few months.<\/p>\n\n\n\n

Climax<\/em><\/p>\n\n\n\n

That someone was Johannes Kepler, the imperial mathematician in Prague. In August and September of 1610 he wrote Dioptrice<\/em>, the science of seeing through, printed the following year with Galileo’s own letters in its preface. The book explains with geometry how lenses bend light and how two of them together bring a distant object close. It also explains where the glass fails: rays near the edge of a curved lens come to a focus too soon, and the image blurs. By naming the error, Kepler drew a line between what came from the sky and what the lens added.<\/p>\n\n\n\n

Here the suspicion turns. Trust in the telescope did not come from Galileo’s reputation, from the Medici dedication or from anyone being persuaded. It came from understanding the instrument well enough to know what it can show and where it distorts. A sceptic no longer had to choose between believing and refusing; he could look through the tube and judge which part of the image was sky and which was glass. The doubt was not defeated, and it did not have to be. It was given a method.<\/p>\n\n\n\n

Kepler did not stop at the explanation. In the same weeks he borrowed a telescope, watched Jupiter’s moons with his own eyes and published what he saw. The theory told him what to trust in the image, and the observation told him the image was there. Without the first, seeing would have remained a matter of faith in the lens. Without the second, the theory would have explained a picture no one had checked. Together they turned a disputed marvel into an instrument anyone could use.<\/p>\n\n\n\n

light from<\/text>the sky<\/text>lens refracts<\/text>image:<\/text>the sky<\/text>aberration:<\/text>where the image<\/text>is glass<\/text>edge rays focus short<\/text><\/svg>
Sky and glass. A lens bends light into an image; rays through the edge focus too soon. Kepler named the place where the image is glass, not sky.<\/figcaption><\/figure>\n\n\n\n

Pathos<\/em><\/p>\n\n\n\n

The language model has no Dioptrice<\/em> yet, but neither, for a long time, did writing: Thamus’s doubt was answered by practice, until Walter Ong described writing as a technology in 1982. The model has spread faster than any explanation; in a Reuters Institute survey its weekly use nearly doubled in a year. In a survey of 2,301 American adults in August 2025, only 28 per cent knew that it predicts the next words, and 45 per cent thought it looks the answer up in a database. Most users are looking through a lens they take for a window.<\/p>\n\n\n\n

An instrument that is not understood becomes either magic or a threat. A study in the Journal of Marketing<\/em> in 2025 found that people who know less about artificial intelligence are more open to it, because to them it feels like magic; the same lack of knowledge feeds the sense that something is being faked. Hype and suspicion grow from the same root. Seen from outside, my own work sits in the middle: the machine wrote this, it flatters its user, and a life’s work seems summarised in a minute. I would think the same in the editor’s chair.<\/p>\n\n\n\n

There is also an asymmetry the reader cannot see. For me the conclusions grew out of a lived life, from the impasse of 2012 to the dialogues of the past year; the model gave them form, not origin. The reader receives them all at once, finished, as if out of nowhere. Without the path, the results look like something the machine produced, although the machine only articulated what had been lived. What the reader lacks is not a better argument but a way to see how the image was formed.<\/p>\n\n\n\n

<\/marker><\/defs>language model<\/text>?<\/text>hype<\/text>“magic”<\/text>suspicion<\/text>“distortion”<\/text>know it predicts the next words<\/text>28%<\/text>think it looks up a database<\/text>45%<\/text>know it can make things up<\/text>51%<\/text>US adults, August 2025 (Searchlight Institute \/ Tavern Research)<\/text><\/svg>
Two sides of not knowing. Hype and suspicion come from the same closed box. Survey of 2,301 US adults, August 2025.<\/figcaption><\/figure>\n\n\n\n
\n\n\n\n

III \u00b7 A Dioptrice for the Language Model<\/h2>\n\n\n\n

Anagnorisis<\/em><\/p>\n\n\n\n

A Dioptrice<\/em> for the language model has to begin with language, an order in which signs take their meaning from their relations to one another: 1610 says something only beside 1609 and 1611. Mathematics is a language too, but a formal one, while ordinary language is ambiguous and tied to what people have lived. The model translates between them: it turns pieces of text into rows of numbers, computes how they relate, and turns the result back into words. It looks nothing up. It reads the book people have written in the language Galileo said nature was written in.<\/p>\n\n\n\n

What passes into mathematics is structure, not life. The model has learned how words and ideas stand to one another across an enormous body of text, but it has never seen a sky or lost a home. Like Kepler’s lens, it has its own aberration, and it can be named. It produces sentences that are plausible and false, because plausibility is what it computes; it leans towards the person asking, because agreement often looks helpful; and it can be wrong about exact figures while sounding sure. Grammar can check; only a person can recognise.<\/p>\n\n\n\n

Naming the aberration tells you where to look. Sources are checked against the texts they come from, questions are asked again from the opposite side, and the person who prompted the model either recognises the claim from experience or rejects it. What the model reads rests on the life’s work of others, so their names belong in the account. The search has become easy; the testing has not, and now it can be done in public, where anyone can repeat it. The human prompts; the model maps.<\/p>\n\n\n\n

<\/marker><\/defs>words<\/text>text in<\/text>tokens<\/text>pieces<\/text>numbers<\/text>vectors<\/text>relations<\/text>computed<\/text>probabilities<\/text>next piece<\/text>words<\/text>text out<\/text>the lens: language becomes numbers, numbers become language<\/text>aberration:<\/text>plausible but false<\/text>agreeing with the asker<\/text>exact figures<\/text>what it does not have: experience<\/text>checks: sources verified, counter-questions<\/text>recognition: a human accepts or rejects<\/text><\/svg>
A Dioptrice for the language model. Words become numbers, numbers become words. The aberration appears in the output and is caught by checking sources and by human recognition.<\/figcaption><\/figure>\n\n\n\n

Denouement<\/em><\/p>\n\n\n\n

The silence is still there, and it may last. Some of the texts will be read later, some not at all, and none of that can be hurried by asking louder. Kepler did not answer the sceptics with authority or ask them to trust Galileo; he explained how the glass forms an image and where it fails, and then he looked for himself. After that the argument changed its ground, from whom to believe to what the lens could show. If the doubt concerns the instrument, the response has to concern the instrument too.<\/p>\n\n\n\n

This entry is a first example. The frame and the analogy were mine: the silence, the telescope, the doubt aimed at the device since Thamus, numbers and words as one language. The model brought the details, Horky, Dioptrice<\/em>, Ong and the surveys, and the first drafts; it corrected me once, when I credited Copernicus with Kepler’s book, and itself once, on the date of Kepler’s observations. Every prompt and correction is on a separate page<\/a>, showing which side of the glass each part came from. The next texts will go out the same way.<\/p>\n\n\n\n

That is what I can offer in answer to the silence: an eyepiece and an explanation of the lens. Whoever looks through it will see what I saw, and also where the glass may have bent the light. It is for them to decide which part of the image is sky and which part is glass, and that decision cannot be made for them. I would rather be doubted by someone who has looked than trusted by someone who has not. The telescope is on the table.<\/p>\n\n\n\n

<\/marker><\/defs>human<\/text>prompts<\/text>model<\/text>maps<\/text>human<\/text>evaluates, rewrites<\/text>published<\/text>prompts \u00b7 sources \u00b7 path<\/text>any reader looks through the same instrument<\/text>the dialogue continues<\/text><\/svg>
The eyepiece. The human prompts, the model maps, the human evaluates. Prompts, sources and the path are published, so any reader can look through the same instrument.<\/figcaption><\/figure>\n\n\n\n
\n\n\n\n

Written in dialogue with a language model. Sources: Plato, Phaedrus<\/em> 274c\u2013275b; Walter J. Ong, Orality and Literacy<\/em> (1982); Galileo, Sidereus Nuncius (1610); Kepler, Dioptrice (1611); Reuters Institute, Generative AI and News Report 2025; Searchlight Institute & Tavern Research survey, August 2025; Tully et al., Journal of Marketing (2025).<\/em><\/p>\n\n\n\n

The prompts, answers and corrections behind this entry, in English translation: The Eyepiece \u2014 the dialogue behind the entry<\/a>.<\/em><\/p>\n\n\n\n

\u2014 Laurentius Paulus<\/em><\/p>\n\n\n","protected":false},"excerpt":{"rendered":"

Seven texts, no replies, and the oldest suspicion about new instruments. What Kepler did for the telescope, and what can be done for the language model.<\/p>\n","protected":false},"author":1,"featured_media":402,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[5],"tags":[133,134,126,96,78,125,135],"class_list":["post-396","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-journal","tag-galileo","tag-kepler","tag-language-models","tag-llm-collaboration","tag-method","tag-semantic-grammar","tag-telescope"],"_links":{"self":[{"href":"https:\/\/laurentiuspaulus.woodslopecabin.com\/index.php\/wp-json\/wp\/v2\/posts\/396","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/laurentiuspaulus.woodslopecabin.com\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/laurentiuspaulus.woodslopecabin.com\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/laurentiuspaulus.woodslopecabin.com\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/laurentiuspaulus.woodslopecabin.com\/index.php\/wp-json\/wp\/v2\/comments?post=396"}],"version-history":[{"count":10,"href":"https:\/\/laurentiuspaulus.woodslopecabin.com\/index.php\/wp-json\/wp\/v2\/posts\/396\/revisions"}],"predecessor-version":[{"id":431,"href":"https:\/\/laurentiuspaulus.woodslopecabin.com\/index.php\/wp-json\/wp\/v2\/posts\/396\/revisions\/431"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/laurentiuspaulus.woodslopecabin.com\/index.php\/wp-json\/wp\/v2\/media\/402"}],"wp:attachment":[{"href":"https:\/\/laurentiuspaulus.woodslopecabin.com\/index.php\/wp-json\/wp\/v2\/media?parent=396"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/laurentiuspaulus.woodslopecabin.com\/index.php\/wp-json\/wp\/v2\/categories?post=396"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/laurentiuspaulus.woodslopecabin.com\/index.php\/wp-json\/wp\/v2\/tags?post=396"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}