The Future of AI Is Helping Us Discover the Past
Historical teaching and research are being revamped by AI.
In February, Google released an upgraded version of its Gemini artificial intelligence model. It quickly became a publicity disaster, as people discovered that requests for images of Vikings generated tough-looking Africans while pictures of Nazi soldiers included Asian women. Building in a demand for ethnic diversity had produced absurd inaccuracies.
Academic historians were baffled and appalled. "They obviously didn't consult historians," says Benjamin Breen, a historian at the University of California, Santa Cruz. "Every person who cares about the past is just like, 'What the hell's going on?'"
Rewriting the past to conform with contemporary political fashions is not at all what historians have in mind for artificial intelligence. Machine learning, large language models (LLMs), machine vision, and other AI tools instead offer a chance to develop a richer, more accurate view of history. AI can decipher damaged manuscripts, translate foreign languages, uncover previously unrecognized patterns, make new connections, and speed up historical research. As teaching tools, AI systems can help students grasp how people in other eras lived and thought.
Historians, Breen argues, are particularly well-suited to take advantage of AI. They're used to working with texts, including large bodies of work not bound by copyright, and they know not to believe everything they read. "The main thing is being radically skeptical about the source text," Breen says. When using AI, he says, "I think that's partly why the history students I've worked with are from the get-go more sophisticated than random famous people I've seen on Twitter." Historians scrutinize the results for errors, just as they would check the claims in a 19th century biography.
Last spring Breen created a custom version of ChatGPT to use in his medieval history class.
Writing detailed system prompts, he generated chatbots to interact with three characters living during an outbreak of bubonic plague in 1348: a traveler passing through Damascus, a disreputable apothecary in Paris, and an upstanding city councilor in Pistoia. The simulation worked like a vastly more sophisticated version of a text-based adventure game—the great-great-great-great-grandchild of the 1970s classic Oregon Trail.
Each student picked a character—say, the Parisian apothecary—and received a description of their environment, followed by a question. The apothecary looks out the window and sees a group of penitents flagellating themselves with leather straps. What does he do? The student could either choose one of a list of options or improvise a unique answer. Building on the response, the chatbot continued the narrative.
After the game, Breen assigned students to write papers in which they analyzed how accurately their simulation had depicted the historical setting. The combined exercise immersed students in medieval life while also teaching them to beware of AI hallucinations.
It was a pedagogical triumph. Students responded with remarkable creativity. One "made heroic efforts as an Italian physician named Guilbert to stop the spread of plague with perfume," Breen writes on his Substack newsletter, while another "fled to the forest and became an itinerant hermit." Others "became leaders of both successful and unsuccessful peasant revolts." Students who usually sat in the back of the class looking bored threw themselves enthusiastically into the game. Engagement, Breen writes, "was unlike anything I've seen."
For historical research, ChatGPT and similar LLMs can be powerful tools. They translate old texts better than specialized software like Google Translate can because, along with the language, their training data include context. As a test, Breen asked GPT-4, Bing in its creative mode, and Anthropic's Claude to translate and summarize a passage from a 1599 book on demonology. Written primarily in "a highly erudite form of Latin," the passage included bits of Hebrew and ancient Greek. The results were mixed but Breen found that "Claude did a remarkable job."
He then gave Claude a big chunk of the same book and asked it to produce a chart listing types of demons, what they were believed to do, and the page numbers where they were mentioned. The chart wasn't perfect, largely because of hard-to-read page numbers, but it was useful. Such charts, Breen writes, "are what will end up being a game changer for anyone who does research in multiple languages. It's not about getting the AI to replace you. Instead, it's asking the AI to act as a kind of polymathic research assistant to supply you with leads."
LLMs can read and summarize articles. They can read old patents and explain technical diagrams. They find useful nuggets in long dull texts, identifying, say, each time a diarist traveled. "It will not get it all right, but it will do a pretty decent job of that kind of historical research, when it's narrowly enough focused, when you give it the document to work on," says Steven Lubar, a historian at Brown University. "That I'm finding very useful."
Unfortunately, LLMs still can't decipher old handwriting. They're bad at finding sources on their own. They aren't good at summarizing debates among historians, even when they have the relevant literature at hand. They can't translate their impressive patent explanations into credible illustrations. When Lubar asked for a picture of the loose-leaf binder described in a 19th century patent, he got instead a briefcase opening to reveal a steampunk mechanism for writing out musical scores. "It's a beautiful picture," he says, "but it has nothing to do with the patent which it did such a good job of explaining."
In short, historians still have to know what they're doing, and they have to check the answers. "They're tools, not machines," says Lubar, whose research includes the history of tools. A machine runs by itself while a tool extends human capacities. "You don't just push a button and get a result."
Simply knowing such new tools are possible can unlock historical resources, permitting new questions and methods. Take maps. Thousands of serial maps exist, documenting the environment at regular intervals in time, and many have been digitized. They show not only topography but buildings, railways, roads, even fences. Maps of the same places can be compared over time, and in recent years historians have begun to use big data from maps.
Katherine McDonough, a historian now at Lancaster University in the United Kingdom, wrote her dissertation on road construction in 18th century France. Drawn to digital tools, she was frustrated with their inability to address her research questions. Map data came mostly from 19th and 20th century series in the U.S. and United Kingdom. Someone interested in old French maps was out of luck. McDonough wanted to find new methods that could work with a broader range of maps.
In March 2019, she joined a project at The Alan Turing Institute, the U.K.'s national center for data science and AI. Knowing that the National Library of Scotland had a huge collection of digitized maps, McDonough suggested looking at them. "What could we do with access to thousands of digitized maps?" she wondered. Collaborating with computer vision scientists, the team developed software called MapReader, which McDonough describes as "a way to ask maps questions."
Combining maps with census data, she and her colleagues have examined the relationship between railways and class-based residential segregation. "The real power of maps is not necessarily looking at them on their own, but in being able to connect them with other historical datasets," she says. Historians have long known that higher-class Britons lived closer to passenger train stations and farther from rail yards. With their noise and smoke, rail yards seemed like obvious nuisances whose lower-class neighbors lacked better options. Matching maps with census data on occupations and addresses showed a more subtle effect. The people who lived near rail yards were likely to work in them. They weren't just saving on rent but decreasing their commuting times.
MapReader doesn't require extreme geographical precision. Drawing on techniques used in biomedical imaging, it instead divides maps into squares called patches. "When historians look at maps and we want to answer questions, we want to know things like, how many times does something like a building appear on this map? I don't need to know the exact pixel location of every single building," says McDonough. Aside from streamlining the computation, the patchwork method encourages people to remember that "maps are just maps. They are not the landscape itself."
That, in a nutshell, is what historians can teach us about the answers we get from AI. Even the best responses have their limits. "Historians know how to deal with uncertainty," says McDonough. "We know that most of the past is not there anymore."
Everyday images are scarce before photography. Journalism doesn't exist before printing. Lives go unrecorded on paper, business records get shredded, courthouses burn down, books get lost. Conquerors destroy the chronicles of the conquered. Natural disasters strike. But tantalizing traces remain. AI tools can help recover new pieces of the lost past—including a treasure trove of ancient writing.
When Mount Vesuvius erupted in 79 C.E., it buried the seaside resort of Herculaneum, near modern-day Naples and the larger ancient city of Pompeii. Rediscovered in the 18th century, the town's wonders include a magnificent villa thought to be owned by the father-in-law of Julius Caesar. There, early excavators found more than 1,000 papyrus scrolls—the largest such collection surviving from the classical world. Archaeologists think thousands more may remain in still-buried portions of the villa. "If those texts are discovered, and if even a small fraction can still be read," writes historian Garrett Ryan, "they will transform our knowledge of classical life and literature on a scale not seen since the Renaissance."
Unfortunately, the Herculaneum scrolls were carbonized by the volcanic heat, and many were damaged in early attempts to read them. Only about 600 of the initial discoveries remain intact, looking like lumps of charcoal or burnt logs. In February, one of the scrolls, a work unseen for nearly 2,000 years, began to be read.
That milestone represented the triumph of machine learning, computer vision, international collaboration, and the age-old lure of riches and glory. The quest started in 2015, when researchers led by Brent Seales at the University of Kentucky figured out how to use X-ray tomography and computer vision to virtually "unwrap" an ancient scroll. The technique created computer images of what the pages would look like. But distinguishing letters from parchment and dirt required more advances.
In March 2023, Seales, along with startup investors Nat Friedman and Daniel Gross, announced the Vesuvius Challenge, offering big money prizes for critical steps toward reading the Herculaneum scrolls. A magnet for international talent, the challenge succeeded almost immediately. By the end of the year, the team of students Youssef Nader, Luke Farritor, and Julian Schilliger had deciphered more than enough of the first scroll—about 2,000 characters—to claim the grand prize of $700,000. "We couldn't have done this without the tech guys," an excited Richard Janko, a professor of classical studies at the University of Michigan, told The Wall Street Journal.
Although only about 5 percent of the text has so far been read, it's enough for scholars to identify the scroll's perspective and subject. "Epicureanism says hi, with a text full of music, food, senses, and pleasure!" exulted Federica Nicolardi, a papyrologist at the University of Naples Federico II. This year the project promises a prize of $100,000 to the first team to decipher 90 percent of four different scrolls. Reclaiming the lost scrolls of Herculaneum is the most dramatic example of how AI—the technology of the future—promises to enhance our understanding of the past.
Editor's Note: As of February 29, 2024, commenting privileges on reason.com posts are limited to Reason Plus subscribers. Past commenters are grandfathered in for a temporary period. Subscribe here to preserve your ability to comment. Your Reason Plus subscription also gives you an ad-free version of reason.com, along with full access to the digital edition and archives of Reason magazine. We request that comments be civil and on-topic. We do not moderate or assume any responsibility for comments, which are owned by the readers who post them. Comments do not represent the views of reason.com or Reason Foundation. We reserve the right to delete any comment and ban commenters for any reason at any time. Comments may only be edited within 5 minutes of posting. Report abuses.
Please
to post comments
Problem is — AI might be able to do all of that.
BUT the foundation is being laid in a time where logic and reason in most academic disciplines is being ignored to promote racial justice.
Even if the idiotic woke bullshit ends — the very foundation of everything involving AI will be corrupted.
Yes, trusting the AI is problematic enough, considering what it’s been trained on and who’s doing the training. Then expecting these same woke cultists to question the results objectively is piling bullshit on bullshit.
The one time I dealt with neural networks, 20 years ago, you were supposed to train the NN on a random half of your dataset, then run the other half through and see if it got the correct answers. All too often, the test run failed miserably.
That example of the three people from 1348 — how many of those students now believe the chats they had were how people actually talked and behaved back them? I don’t mean literally word for word, but the general tone and substance. Does that professor actually believe it?
Yes they believe the ai. They are that retarded
As you mention, AI is subjective. It is based on the training data and the accuracy of trainers. We know right now they outsource the training data to foreign labor. They train the data with a subjective bias. Then on top of that they add rules to add more subjectivity.
AI is essentially training probabilitistic outcomes. That’s it. So if someone chooses to lie to the AI, AI will also lie.
Now we hear they are training data with academic material from the last few decades. Training it on bullshit like 1619 project. Never did they consider training it on primary sources or information closest to the source.
As someone who has used AI and Machine Language for a decade, the faults are obvious. Yet every engineer who comes in recently wants to only work with AI so I spend a lot of time explaining to them why AI has faults and they have to understand those faults to make an algorithm usable.
AI is just as trustworthy as “The Science ”.
In short order, the usual suspects will start calling it “The AI” and that we should trust it.
Vesuvius challenge? They should have used that money to start an academic journal. The basis of the corpus of the common shared human knowledge!
After reading the source of the retarded quote, I was able to determine that it was part of a PR campaign to push for the regulation of AI. One of the PR campaign’s goals is to force corporations to share their algorithms and source data with researchers and by extension, one would assume, academic journals. They specifically cite AI’s inability to prove reproducibility and verifiability while linking to this article: Stop talking about tomorrow’s AI doomsday when AI poses risks today
They want more regulations imposed more quickly by a more diverse group of whiners. In this case, people who make a living in soft, pseudo, and government regulated sciences; people who have to publish to academic journals.
They do throw a bone to “marginalized” communities.
Japan comes out swinging after Joe calls them xenophobic.
It does not appear our allies bought Kirby’s excuse. “It’s not that we are xenophobic,” Sanseito party member Sohei Kamiya told reporters. “We are being cautious after seeing your failures.” “I hope that President Biden will solve the problem in New York before he says things like this,” Nippon Ishin no Kai party member Mizuho Umemura added.
https://www.washingtonexaminer.com/opinion/2990219/bidens-xenophobic-gaffe/
Glad the adults are back in charge and everyone respects us under Joe.
Remember when Trump visited Japan and he copied the prime minister by dumping some koi food in the pond and Jeff and the media pretended it was the biggest diplomatic offence ever?
That was bad. Calling the Japanese racist is small beer compared to that.
You’re forgetting that he committed the greatest crime of all: Being Donald Trump.
We need a wet mop and Grabber Of Pussy Butthurt Salve on aisle 45…
We need a “Seig Heil” and a copy of the Constitution for the elderly gentleman in aisle 88…
That was almost funny. Almost.
Better luck next time Hankie.
I’ve noticed a great influx of use of racist and xenophobic to those pushing back against the globalist open borders policies. Even here.
Trudeauisms 101. If someone disagrees with you call them a “right-wing extremist”.
Which is totally different than saying anyone who disagrees with you is a leftist fascist Marxist leftist extremist leftist.
Not as bad as George HW Bush vomiting on Kiichi Miyazawa.
Yes having the flu at a dinner party is miles worse than calling your allies bigots.
I’ll take being called a bigot by a senile old fool over getting puked on.
Academic historians were baffled and appalled. “They obviously didn’t consult historians,” says Benjamin Breen, a historian at the University of California, Santa Cruz. “Every person who cares about the past is just like, ‘What the hell’s going on?'”
Said the idiot who’s been paying literally no attention for the last ten years.
Are educated people really this dumb?
Yes they are. I worked at a company which sold products to the legal world. They interviewed a guy with a Masters in aeronautics from MIT, who somehow thought we were going to put him through Stanford for a PhD in aeronautics. He was astounded that I, a mere college dropout, knew of the advancing blade problem with helicopters (wing lift varies by the square of the airpseed; if rotor tip speed is 200 mph on a helicopter traveling at 100 mph, one side is generating 9 times the left of the other; and then you’ve got sideways air at the front and back. It gets messy. And pedants: buzz off.).
My firm has had terrible luck with PhD from MIT, Harvard, or other “elite” colleges. They can’t actually build a product, have a sense of entitlement, and wanted the company to basically fund their doctoral research instead of customer contracts. It is to the point we almost never even interview people from those schools.
“Are educated people really this dumb?”
Credentialed, not educated. Universities got out of education decades ago.
Unverifiable sockpuppets are.
Like the shrikes? Say Hank, what happened to your old account?
I hope they soon put AI to use translating the Sumerian tablets. That seems like it would be pretty straightforward.
a disreputable apothecary in Paris
>”Machine learning, large language models (LLMs), machine vision, and other AI tools instead offer a chance to develop a richer, more accurate view of history. AI can decipher damaged manuscripts, translate foreign languages, uncover previously unrecognized patterns, make new connections, and speed up historical research. As teaching tools, AI systems can help students grasp how people in other eras lived and thought.”
Existing systems can either already do that – and do it without ‘hallucinating’ (where ‘AI’ simply makes stuff up) – or its stuff that, no, AI can’t do.
Maybe in 10 years. Maybe. But not now. Right now, AI is not ‘revamping historical research and teaching’. Like, not even a little bit.
Right now it’s saving me two or three hours editing time a week, which is all I really need it for.
>Historians scrutinize the results for errors, just as they would check the claims in a 19th century biography.
How does the historian scrutinize the results for error if the historian *doesn’t already know what the errors are*? What is the AI doing here? Because it seems you set an AI to do something – then you have to go over the something and basically duplicate the work to check for errors. You’re doing *more* work than if you had just done it yourself.
>Writing detailed system prompts, he generated chatbots to interact with three characters living during an outbreak of bubonic plague in 1348: a traveler passing through Damascus, a disreputable apothecary in Paris, and an upstanding city councilor in Pistoia. The simulation worked like a vastly more sophisticated version of a text-based adventure game—the great-great-great-great-grandchild of the 1970s classic Oregon Trail.
So . . . like several other groups have – like the Catholic Church – only for those chatbots to go rogue. Because they’re *not simulations of people living in those eras*, they’re predictive text generators.
Secondly, the people using that chatbot, well, they’re not historians, are they? So they’re not going to have the claimed elevated levels of skepticism that historians have (citation needed – there are historians that swear Cleopatra is black or some woman buried with a weapon must have been trans).
> Engagement, Breen writes, “was unlike anything I’ve seen.”
But what did they *learn*?
>After the game, Breen assigned students to write papers in which they analyzed how accurately their simulation had depicted the historical setting. The combined exercise immersed students in medieval life while also teaching them to beware of AI hallucinations.
Nothing they didn’t already know, apparently. Instead it was an exercise in training them to correct LLM hallucinations, not learning history.
So, as far as I can tell from this article, AI is just being used as a ‘natural language’ interface. Which already exist. This just seems to be an easier way to create those.
Of course, coming with the ever-present danger that your AI interface will just flat out lie to you.
“”Of course, coming with the ever-present danger that your AI interface will just flat out lie to you.””
Pretty sure that’s a feature for governments.
Rilly? Ask the artificial personbot:
What U.S. minor party party first added a women’s suffrage plank, and when?
Or try this: On what date in 1872 was the Civil War income tax repealed?
Artificial Idiocy is plausible denial for memory holes with built-in Alzheimer’s.
Looks like yet another subject Grandfather Hank knows nothing about yet is determined to tell you his opinion on.
Here you go you old retard, fresh from ChatGPT:
“The first U.S. minor party to add a women’s suffrage plank to its platform was the People’s Party, commonly known as the Populist Party. This inclusion happened in 1892 when the party was established, reflecting its progressive stance on various social issues, including women’s rights.
Regarding the Civil War income tax, it was repealed on July 1, 1872. This tax was first imposed during the Civil War to support the Union’s war effort and was a significant source of federal revenue during that time. Its repeal marked a return to the norm of relying on tariffs for federal income.”
Me: “A senile idiot stated the following, what is your response:
“Artificial Idiocy is plausible denial for memory holes with built-in Alzheimer’s.””
ChatGPT: The statement “Artificial Idiocy is plausible denial for memory holes with built-in Alzheimer’s” seems to be a critical and metaphorical comment on artificial intelligence (AI). It suggests that AI can serve as a convenient excuse—or “plausible denial”—for failures in memory or information management, likening these failures to “memory holes” or gaps in memory associated with conditions like Alzheimer’s disease.
The comment could be interpreted as a skepticism about AI’s reliability and effectiveness, implying that sometimes AI might fail in ways that mimic human forgetfulness or errors, but these failures could be excused as limitations of the technology rather than oversight or intent. It’s a provocative way to question how much we should rely on AI and what the implications are when it doesn’t work as expected. Whether one sees this as a witty critique or an exaggerated dismissal might depend on their views of technology’s role and reliability.
So now we will have archeologists AND ai inventing new dinosaurs that never lived. Sweet.
What archeologists are inventing dinosaurs? Which dinosaurs never lived?
At least our extrapolation of missing data could be fiction. Megalodon as far as we know might have been just a silly little 2 ft long shark with a giant mouth, being mostly cartilage we only have the jaw. They invented the brontosaurus, then un-invented it again, and then invented it again, then un-invented it again, then invented it again.
AI systems can help students grasp how people in other eras lived and thought.
WTF? AI struggles with how people in this era live and think.
Instead, it’s asking the AI to act as a kind of polymathic research assistant to supply you with leads.
I don’t think somebody understands what the words “polymath” or “assistant” mean.
Sounds, at best, more like a Rain Man-esque savant than polymath.
It is hilarious that they’ve re-named “Completely Fucking It Up” to “Hallucinating.”
No, no, that’s not a pile of shit, it is “unrealized nutrients.”
Elsewhere I wrote a couple of months back:
The first problem is the framing. Once you use the term “Artificial Intelligence”, people assume that we’re dealing with intelligence. We’re not. What we’re dealing with, at best, can be called “Artificial Idiot Savantry”. Idiot savants are by definition stupid but have high specific functionality and processing power. They lack a high “g factor” , that is, generalised intelligence.
Then there’s “instant expert” syndrome – people who have never considered AI until about five minutes ago are now supposed to be experts given their current position in their firms, or are purporting to be experts and represent themselves as such either internally or to potential clients. You want an expert on AI? Ask anyone who’s been reading science fiction for fifty years. We’ll have read more about it, and thought more about it than anyone else except a long-time AI researcher. There’s an entire subcategory of SF that is, in effect, AI thought experiments.
I think it was J.E. Gordon who, in “Structures, Or Why Things Don’t Fall Down”, identified a three-generation cycle in new civil engineering technology, like bridge design and materials. In my words, A. We don’t know much, so let’s really overengineer this. B. We think we get it, so we don’t need too much of a safety margin now. C. We really understand it now, and everyone before us was too conservative. Disaster follows thereafter.
A similar risk is likely with respect to generative AI and its use in corporations. The first generation will use it as a specific tool, have human firewalls, and be very cautious. The second generation will broaden its applications, reduce human involvement, and be prudent. The third generation will continue with broad applications, will let AIs “manage” other AIs, and will be overconfident and hence careless. The major difference between bridges and AIs, in context, will be the length of the generations. For bridges it might be 20 or more years, for AIs, it might be a matter of a year or two.
Every new technology eventually leads to a disaster. Bridges, airplanes, etc.
Hopefully we aren’t all flying inside this newly designed plane at the same time when it crashes.
I’d add “financial products” to that list.
Is it helping us “Discover the past,” or is it inventing a version of the past that nobody will be able to verify?
The prolog to AI’s sudden escape from inanity was the web’s transformation of how scholars train themselves.
Until the turn of this century, nobody enjoye really d deep knowledge of the historical record before the invention of printing, because manuscripts were hard to access and harder to read. Then, all of a sudden, character recognition software transcribed the paper chase archives of the fifteenth through ninteeth centuries , and shazam, Big History surfaced online.
All of a sudden archival research seemed a lot easier than it had been, just as today it is dawning upon us that language is not quite as complex as it seemed a decade ago.