Back to the 1830s.
Generative AI photos and video are becoming so good, it is becoming hard to differentiate from ‘real’ imagery. I have been wondering lately whether we might have to go backward to move forward. Before photography became a common method of recording the world. Before movies, television, the internet, smartphones, and the torrent of imagery that now occupies so much of our waking consciousness. Before we became accustomed to believing that seeing something meant it actually happened.
Of course, photography was already being invented. Nicéphore Niépce had succeeded in producing permanent photographic images in the 1820s, but Daguerre’s practical process wouldn’t be publicly announced until 1839.
For most people, the world was still encountered through words, paintings, drawings, engravings, correspondence, and firsthand experience. A newspaper described an event in detail. A traveler wrote romantically about a foreign landscape. An artist depicted a battle in different media. A witness related what had happened.
The reader had to imagine. And, importantly, the reader had to decide what to believe.
Not that people were necessarily better at discerning truth in the 1830s. Quite the contrary. America’s newspapers were fiercely partisan, sensational, frequently unreliable, and perfectly capable of circulating outrageous falsehoods. We don’t generally learn this in American public schools. Not even in college, in my experience.
Consider Founding Father Thomas Jefferson as an example.
By 1830, Jefferson had been dead only four years. He was already becoming something of an American icon, though those who had lived through his presidency remembered a considerably more complicated man.
During the election of 1800, Federalist newspapers and clergymen savaged him for his religious beliefs. Jefferson, whose religious philosophy departed substantially from orthodox Christianity, was portrayed as an infidel, an atheist, even a threat to the moral foundations of the nation.
Sound familiar? This was happening against the backdrop of the Second Great Awakening, a period of religious revival that would profoundly influence American culture.
Jefferson, whose religious beliefs were broadly deistic, the advocate of religious liberty and separation of church and state, began attending worship services in the House of Representatives in 1802. His attendance was reported in newspapers, helping to counter accusations of irreligion. Historians caution that this was not necessarily an insincere performance; he also believed religion could serve a useful social purpose.
Still, it is a revealing moment. A president’s personal beliefs, public image, and political fortunes were being negotiated through the media of his day.
Then there was the 1807 treason trial of his former vice president, Aaron Burr. Chief Justice John Marshall ruled that Jefferson could be subpoenaed to produce evidence. Jefferson did not appear to testify in person, and he supplied only some of the requested material. Having publicly declared Burr’s guilt before the trial, he was hardly an impartial observer.
Jefferson was brilliant, certainly. The author of extraordinary declarations about liberty. A formidable intellect. He was also politically calculating, contradictory, and capable of immense moral blindness. The man who proclaimed human equality enslaved other human beings.
Just a man. Like any other.
Which Jefferson was real? The one described by his enemies? The one celebrated by his admirers? The one immortalized in portraits and monuments? Or the considerably more complicated person revealed by correspondence, public records, and historical scholarship?
There was no photography necessary to manufacture competing representations of Jefferson.
Only language. And an audience willing to believe.
In 1840, the first useful national census figures suggest that roughly 78 percent of American adults could read and write, rising to about 91 percent among white adults. Today, basic literacy is far more widespread, but recent federal assessments indicate that more than a quarter of working-age adults struggle with important aspects of reading proficiency. We have vastly expanded access to information without necessarily ensuring our ability to evaluate it. And now we’re introducing technologies capable of manufacturing persuasive information faster than anyone can reasonably investigate it.
The camera didn’t invent deception. It introduced something different: a new kind of evidence. For the first time, a scene could be recorded through the physical action of light rather than reconstructed entirely through an artist’s hand or a writer’s description. A photograph appeared to offer a direct encounter with reality. And for nearly two centuries, we’ve increasingly treated it as such.
Of course, photographs have never been entirely truthful. Cropping, staging, selective framing, retouching, darkroom manipulation, and outright fabrication have accompanied photography practically since its beginning.
The camera records only what it is pointed toward. The photographer decides what the viewer will see, and perhaps more importantly, what the viewer will not. Often a photograph is opinion, not fact. A feeling enhanced by what is not included. But there was usually something physically present, somewhere, at some moment, reflecting light into a camera. A witness of sorts.
Today, generative artificial intelligence can produce a convincing photograph without any such encounter taking place.
No camera. No photographer. No location. No moment. Just computation and an image that looks entirely plausible.
The implications are staggering.
It’s not merely that we may believe things that never happened. It’s that we may begin disbelieving things that actually did.
The genuine photograph, the authentic recording, the eyewitness document … all become vulnerable to dismissal. Anything inconvenient can be accused of being generated. Anything emotionally satisfying can be accepted as real. And when everything can be manufactured, everything can be doubted.
That’s a peculiar kind of victory for falsehood. As a photographer, I find this deeply unsettling.
I’ve spent years looking for light, selecting compositions, waiting for moments, traveling to places, sometimes enduring considerable inconvenience to make a photograph. The photograph is an interpretation, certainly. It reflects my choices, my sensibilities, my limitations.
But I was there. I saw something. I made a record of it. That distinction matters to me.
And as a writer, I have an equally uncomfortable realization: language isn’t inherently more truthful than imagery. Jefferson’s contemporaries demonstrated that rather effectively.
A beautifully composed paragraph can mislead just as convincingly as a fabricated photograph. A confident explanation can conceal omissions, stereotypes, selective history, and unsupported assumptions.
We’re surrounded by people explaining the world to us. Talking heads, influencers, commentators, experts, self-appointed experts, and now artificial intelligences capable of producing endless authoritative-sounding explanations.
So many explanations. So few footnotes. And even fewer admissions of uncertainty. Those talking heads, smirk on the face, eyes narrow with purpose, glowing with falsehoods, being compensated generously for exponentiating them.
What troubles me isn’t simply the quantity of misinformation. It’s the growing expectation that every complicated subject can be reduced to an easily digestible conclusion.
Nuance is inconvenient. Discernment takes effort.
Doubt doesn’t make for compelling clickbait.
And apparently the three most intellectually honest words in the language … I don’t know … have become something of a commercial liability.
Umberto Eco understood something important about communication. The person controlling the message doesn’t necessarily control its meaning. The recipient brings an entire world of experience, knowledge, prejudice, and expectation to the act of interpretation.
What happens, then, when our technologies begin learning those expectations?
When messages can be shaped to fit the assumptions of their recipients? When an explanation can be personalized not merely to inform, but to persuade? When the machinery delivering information can also discover which version of that information we’re most inclined to accept? When the machinery can also guarantee the message lands with whatever code the recipient needs to align with the proper interpretation?
The partisan newspapers of Jefferson’s America sought to influence their audiences through ideology, rhetoric, selective reporting, and appeals to religion or patriotism.
Artificial intelligence offers the possibility of doing something more sophisticated: adjusting the message to the individual, potentially learning from each interaction how to make the next message more persuasive.
The old propagandist had to construct an argument for an audience. The new propagandist may be able to construct a different argument for every person in that audience. Unique dog whistles for every individual American, cheap and easy.
That doesn’t guarantee persuasion. Human beings remain capable of questioning and rejecting what they’re told. But the distance between delivering a message and influencing its interpretation may be narrowing.
This is where the problem becomes much larger than fabricated photographs. We’re no longer concerned only with whether a message is authentic. We’re concerned with whether the process by which we interpret messages remains independent.
And perhaps that brings us back to the 1830s. Not to the horse and buggy, the candle, the quill pen, or some imaginary golden age of intellectual honesty.
1830s America was not a more truthful place. Its citizens could be misled by partisan newspapers, political mythology, religious prejudice, and the confident declarations of prominent men. Many were denied the opportunity to participate in public life at all. But the absence of photography reminds us of something we may have neglected.
Before we could rely on the apparent authority of a mechanically produced image, we encountered much of the wider world through personal accounts of it. We read descriptions. We compared testimony. We heard contradictory narratives. Sometimes we challenged them. Often we didn’t.
Perhaps what we need to recover isn’t the past itself, but the discipline its limitations could demand.
Reading slowly. Reading skeptically. Reading competing accounts. Examining sources. Considering whether a writer has actually established a claim or merely expressed it persuasively. Relearning the difference between witnessing an event and being shown a representation of one. Realizing that each witness can have completely different perceptions, recollections, and interpretations … and only our commonsense and education can guide us to theory, if not fact. Demanding provenance from photographs, references from writers, and evidence from people who confidently announce that they know how the world works.
Above all, becoming comfortable with uncertainty.
There are technological efforts underway to authenticate digital images and document their origins. They may help. But even a perfectly authenticated photograph cannot tell us everything outside its frame. And knowing who created an image doesn’t guarantee that the image tells the truth.
Technology can assist verification. It cannot relieve us of judgment.
I don’t want to abandon photography. Far from it. Nor do I want to surrender the extraordinary possibilities of digital communication, or pretend that artificial intelligence has nothing valuable to offer. But I do want to recover something we’ve been steadily giving away.
The responsibility to interpret. The willingness to question. The intellectual patience to distinguish what we know from what we merely believe. And have the integrity to stand behind those beliefs.
We have been constructing persuasive representations of reality for centuries. First through language, then through mechanical images, and now through systems capable of manufacturing both.
Each technological advance has changed the means of persuasion. None has eliminated our responsibility to think.
Back to the 1830s, then. Not because the past was more truthful. But because the future may demand that we learn to read and judge the world for ourselves again.
If this makes you think of independent blogs, and our early ‘semiological guerilla warfare’ before the advent of social media services, then my work is done here.
Notes and references
-
The Metropolitan Museum of Art. Daguerre (1787–1851) and the Invention of Photography. On Niépce’s pioneering photographic experiments and the development of the daguerreotype.
-
Library of Congress. Election of 1800. Primary-source material documenting the intensely partisan presidential campaign and the political maneuvering surrounding it.
-
Library of Congress. Religion and the Federal Government, Part 2. Documents Jefferson’s religious controversies, his attendance at church services in the Capitol, and the relationship between religion and politics.
-
Federal Judicial Center. The Aaron Burr Treason Trial. A historical account of the 1807 proceedings, including Chief Justice Marshall’s ruling on presidential subpoenas.
-
National Constitution Center. Aaron Burr’s Trial and the Constitution’s Treason Clause. On Jefferson’s involvement and the constitutional issues arising from the prosecution.
-
U.S. Census Bureau and National Center for Education Statistics. Historical Statistics: Illiteracy in the United States; PIAAC 2023: Highlights of U.S. National Results. The 1840 census reported illiteracy among 22% of Americans aged 20 and older, and 9% of white adults. In 2023, 28% of Americans aged 16–65 scored at Level 1 or below in reading proficiency. The historical figures measure basic literacy, while the modern assessment measures reading comprehension and proficiency; the two are not directly comparable.
-
Umberto Eco. Towards a Semiological Guerrilla Warfare (1967), collected in Travels in Hyperreality. On communication, competing interpretive codes, and the recipient’s role in constructing meaning.
-
Coalition for Content Provenance and Authenticity. C2PA Specifications. On digital content provenance, authentication, and their limitations.