> Dan, who created the video of Pompeii, says he recognises a lots of details in his videos are historically inaccurate.
"AI-generated content isn't perfect, and while I strive for accuracy, these videos are more about evoking the feeling of a time period rather than being a 100% factual recreation.
"They're more like artistic interpretations rather than strict documentaries."
As someone technical who works in creative circles, I hear this kind of reasoning all the time and it drives me nuts. The “it’s just art bro, don’t take it so seriously” excuse is so tired, but like the best fake excuses there’s no way to counter it.
Well that's plainly not true. Striving for accuracy is going through the script and doing this thing called "fact checking". It's an actual job that used to mean something to writers, journalists, documentarians...
The noble goal of egalitarian democratization of media has had the unintended consequence of completely blurring the lines between quality and ... non-quality information. What galls me most is the people who care about fact and truth, and used to be the guardians or advocates for such things, are just drowned in a sea of bots, malicious actors, and the ignorant and are thereby discouraged.
>What galls me most is the people who care about fact and truth, and used to be the guardians or advocates for such things, are just drowned in a sea of bots, malicious actors, and the ignorant and are thereby discouraged.
Some people have always cared quite a lot about the truth, but major media cared in part because lawsuits over libel, slander, or defamation are expensive.
However, there’s little point in suing 1 of 10,000 people all spouting nonsense. At least until they gain a little traction and thus enough resources to be worth suing like Alex Jones etc.
Bullshit. The world was never made of saints, but there were real institutional checks and balances (asking the accused for comment, publishing retractions, institutional fact-checkers next to editorial staff) and a far greater cultural sense that truth was valuable and lies were shameful. These are easy to appreciate, warts and all, now that they are gone. They leave a big hole.
There was one good innovation in factuality to come out of the social media revolution, but it has drawn the ire of its owner and looks like it will be put down shortly.
One of my favorite quotes (slightly edited for brevity and and readability):
------
It is a melancholy truth that a suppression of the press could not more completely deprive the nation of it’s benefits, than is done by it’s abandoned prostitution to falsehood. Nothing can now be believed which is seen in a newspaper. The real extent of this state of misinformation is known only to those who are in situations to confront facts within their knowledge with the lies of the day.
I really look with commiseration over the great body of my fellow citizens, who, reading newspapers, live & die in the belief that they have known something of what has been passing in the world in their time: whereas the accounts they have read in newspapers are just as true a history of any other period of the world as of the present, except that the real names of the day are affixed to their fables.
General facts may indeed be collected from them, such as that Europe is now at war, that Bonaparte has been a successful warrior, that he has subjected a great portion of Europe to his will etc. But no details can be relied on. I will add that the man who never looks into a newspaper is better informed than he who reads them; inasmuch as he who knows nothing is nearer to truth than he whose mind is filled with falsehoods and errors. He who reads nothing will still learn the great facts, and the details are all false.
Thomas Jefferson - 1807
------
That would have sounded rather hyperbolic about 30 years ago, perhaps even more recently. But now it sounds just about right. I think what many do not realize is that the past ~70-80 years in the developed world, or in other words the complete living memory of just about every person alive today in those regions, was atypical in just about every single way - including some effort towards the pursuit of objective truth. We're not entering some scary unknown place, we're simply returning to the status quo that existed throughout the breadth of human civilization.
This also makes reading the ancient philosophers somewhat literally unbelievably insightful. Plato oft sounds far closer to a proven prophet than a philosopher when you read his writings on, for instance, politics!
> That would have sounded rather hyperbolic about 30 years ago, perhaps even more recently. But now it sounds just about right. I think what many do not realize is that the past ~70-80 years in the developed world, or in other words the complete living memory of just about every person alive today in those regions, was atypical in just about every single way - including some effort towards the pursuit of objective truth.
https://en.m.wikipedia.org/wiki/Gell-Mann_amnesia_effect
>> The Gell-Mann amnesia effect is similar to Erwin Knoll's law of media accuracy, which states: "Everything you read in the newspapers is absolutely true except for the rare story of which you happen to have firsthand knowledge."[3]
That cite says it's from 1982.
So no, the news of 40+ years ago wasn't particularly accurate.
This idea that fact checking is preferable or possible or even likely to happen seems wrong. The answer isn't to fix the disease that is Social Media but to inoculate ourselves from it entirely.
Its either that or yield to the idea that human psychology is so malleable that global corporations (and the engineers who make millions working for them) control us like marionettes.
By that do you mean stop using it entirely? Because I don't think it's possible for many (maybe most) people to engage without being affected—at least with the currently prominent social media formats.
I think there are giant weak spots in the product offerings of social media companies that provides opportunities for regional alternatives to win. For example, imagine an utterly transparent social media website with low operating costs that could offer services to a region at minimal or subsidized cost.
I don't think social media requires or should be world-scale. I think its hubris to think Facebook etc are meaningfully connecting us to individuals around the world when our social graphs our overwhelmingly local and sequestered.
Finally, because nobody ever asks this question before they build it, there is a potential for regional networks to be more harmful - for example in a region with active genocide you could see regional networks aggravate that harm more then the US tech companies do.
It's a tough nut to crack but I don't think it compares to the hardest challenges humanity has already conquered.
>I think there are giant weak spots in the product offerings of social media companies
Agreed. I'm not sure that the fix is necessarily to regionalize though. I think rather that the weakness is in the format, mechanics, algorithms, etc. We need to re-imagine social media in general (yes, I know this might sound hand-wavy). The true fix might be so different that we don't even call it social media, and...
>there is a potential for regional networks to be more harmful
Exactly. That's why I believe we need a more fundamental change. All of the current formats have the capacity to wreak havoc, whether globally or locally.
>I think its hubris to think Facebook etc are meaningfully connecting us to individuals around the world
Yeah, I don't know that Facebook was ever meaningfully trying in this regard, except to promote some tag line, "something-something connecting the world" to make it seem they were on some noble quest. All they really wanted was eyeballs, and judging from the sheer stream of garbage that is today's FB feed, they're not even pretending anymore.
>the idea that human psychology is so malleable that global corporations (and the engineers who make millions working for them) control us like marionettes.
But it is that malleable, who would even dispute that?
Yes, that's why the biggest spender always wins elections. And why heavily marketed movies never flop.
But the way these discussions always seem to go, is that anyone who provides counterexamples is immediately told they're imagining things and/or gets insulted for being unable to see what's really going on.
Oh of course you can counter it. The issue is less with artistic interpretation but with the doublethink happening here. You can depict something as a stylized artistic representation that does not pretend to be accurate. I'm pretty sure the bombing of Guernica did not look exactly as Picasso portrayed it, but his painting is still a powerful piece that we can use to understand the artist's feelings about the historical event. And even in the most rigorous of documentaries, there is still room for artistry: you can convey information in an interesting and thoughtful manner that doesn't sacrifice factuality.
You can't really simultaneously say that you strive for accuracy but, at the end of the day, you're "just" creating an artistic interpretation. That seems to conflate artistry with a failure to capture historical accuracy, which is so bizarre since those things are more or less orthogonal to each other.
I think we can safely call a spade a spade here: people making lazy AI historical videos are just that; lazy. They found a cheap and easy way to churn out content that makes money, and their greed outweighs their commitment to truth or historicity.
Just seems to be one more front in this War on Truth we've found ourselves in over the last decade or so. Yes, I know some will say it's been longer than that, the MSM is evil, etc. It's true that the 24-hour news cycle and the replacement of news with infotainment, purveyance of propaganda as news by some outlets, etc. have not been helpful. But, about a decade ago, we entered a completely new phase that is qualitatively different. And it feels like the end stage.
Whether it's for political gain or profit or otherwise, the value of truth and shared reality has been greatly diminished. It's not just that the videos aren't accurate and the creators are so dismissive. It's not even as much about the supply side any more.
It's that half or more of the population is so willing to choose their own reality and could care less whether it's actually true.
> The “it’s just art bro, don’t take it so seriously” excuse is so tired
There’s nothing wrong with it as long as they’re not presenting it as history but are presenting it as the work of fiction it is.
Unfortunately they’ll, at best, have fine print that says it’s not history but the whole presentation is supposed to encourage the viewers to believe this ks real.
Hollywood has been creating historically inaccurate portrayals for over a hundred years. I'm really not seeing the harm in these shitty little vignettes. If anything, it's more concerning that TikTok's algorithms are able to get people to give their attention to such poor quality content.
> I'm really not seeing the harm in these shitty little vignettes
Pompei and the Black Death aren't historical events that continue to evoke grievances today.
But creating an AI generated recreation of the Nakba or Nanjing Massacre can evoke extreme reactions to this day, and actually instigate diplomatic fallout and even violence in the modern world.
If you keep percolating and percolating such content, then it can complete shift narratives.
Technology is inherently neutral - it's us humans who use technology for good and for bad - but moderation has simply not kept up the rate of change.
> it's more concerning that TikTok's algorithms are able to get people to give their attention to such poor quality content
People will continue watching what they want, and recommender algorithms will continue to recommend them as such.
If recommender algorithms are not regulated, then this is the status quo that arises.
On that front, statist countries like Singapore have better managed these problems, but they also pair it with permissive national security legislation.
The problem is in economics and scale. Hollywood productions were pretty expensive, hence, there were few of them – and they could be tracked as known narratives. ("Ah, well, this is Ben Hur. We know this.") Gen-AI content is incredibly cheap and thus potentially ubiquitous, and has a real potential to become normalized in common imagination.
On the other hand, I think this makes it even clearer that it's easy to manipulate the truth, since now more people can do it. Before, it was only a small subset of the population who had the ability to, and they were more easily believed.
> Despite the videos offering millions a window into history, a number of historians have shared concerns about the accuracy of the content and whether AI can truly resurrect the past, or are we just seeing a polished, modernised version of history designed for engagement?
While I find it annoying to present inaccurate details as "history" I still think we need to put the onus on information consumers to think critically and go beyond taking everything at face value. The world is full of manipulation, lies of omission, and actual lies as well. Taking things at face value (especially on TikTok and YouTube) is foolish. If you don't want to be a fool the ball is in your court, the world will not be becoming more foolproof.
Are you serious? That's like saying expecting people to stay healthy while eating every meal out of a garbage dumpster is laughable. The solution isn't to change the content, it's to stop eating out of a garbage dumpster. And yes, I think you should exercise the self control to not overconsume things, whether it's information or ice cream. If you bombard yourself with more information than you can critically process then that's a "you" problem.
It is dangerous. In most of third world, far right politics is largely based on settling historic scores. To that end, Im guessing this will be weaponised by lots of groups to create content to target minorities.
I do believe bulk of the use for this more than for education purposes, will be for propaganda, radicalisation etc.
While the makers might get away making 'its just art' excuse. This will be a new round of fake news/history flood unseen like ever before. Every bit of vague/abstract statement, or even an absent detail in history books can be taken out of context, or just made from fiction to create narratives against vulnerable groups of people.
One wonders how this will be regulated or if it can even be regulated.
When people talk of AI safety, more than terminator/skynet style apocalypse I largely imagine these sort of things to cause more short/middle term damage.
> In most of the third world, far right politics is largely based on settling historic scores. To that end, I'm guessing this will be weaponised by lots of groups to create content to target minorities.
Current events would suggest that the word "third" is redundant.
Forgive me taking my oafish American hammer to this particular nail, but - this is speech and should not be "regulated," full stop.
The solution to this being used to disseminate falsehoods is education, both about the technology (to make it easier for the average person to spot AI content which at least for the time being is still possible) and about topics likely to have falsehoods spread about them.
Isn't education also "just speach" that must not be regulated then?
If not, then I believe anything that present history is education, be it on YouTube or in a classroom.
I believe this not because of some preconception or ideology, but because I observed that non classroom material is often referenced by people looking for historical justifications.
I don't really see how this is any different from a videogame or tv show set in some historical time period - it's clearly fiction and there are always historical inaccuracies, often egregious (a newspaper in Gladiator? )
So people are doing historical fiction in video format? That's kinda cool, even if the people doing it aren't experts being meticulous about every last detail.
The majority of the population, including I'd say the majority of people on this site, think historical dramas are accurate. They may not phrase it as such but they implicitly believe things on the screen.
Is that not one of the main benefits that AI is supposed to bring though, and the way that that word is used in that context? Taking things that were previously the ___domain of experts who actually knew what they were doing and allowing the entire general public full ability to participate?
Also why is it suddenly a good thing that people who know nothing can talk shit about it? Is the bar lowered so much already that “can type words in an LLM” is sufficient?
If you spend public funds on NPR or the BBC to produce an educational resource with these kind of problems where the core message was historicity, not some other educational outcome, you'd be wasting national funds and be pilloried for it.
It's contextual. Horrible Histories can make boners. some soft-ed deliverable can make boners about things, not core to the syllabus. If you call it "history" and you want to teach lived experience, then don't put tomato on the roman pizza. It's actually not that hard to get right btw. Mostly, its lazy.
"I Claudius" is perhaps my favorite series of all time, produced by the BBC during its golden age. You might argue the core was not historicity, but then again it'd seem odd to hold the BBC to a lesser standard than a TikTok video for historical recreations. But it's extremely informative in many ways and also extremely misleading to plainly false in others. It's simply the nature of historical recreation.
It did things such as rather dramatically change the widely recorded historical personality/interests/etc of Claudius, presumably to make him a more likeable 'protagonist'. And they weren't pilloried for it whatsoever but rather drowned, in praise and awards.
I Claudius was made in 1976 by BBC entertainment. It might have screened in schools but that was not it's primary purpose. But, carried into the modern era I am pretty sure historians would mutter about the many solecisms. None the less it was better than AI.
Much of the history about roman times is under reconsideration in the light of things like new digs in Pompeii. People projected current social values onto what they found, DNA analysis now permits much stronger indications of filiality or patriality of the remains. "oh, this is a mother protecting her child" can turn out to be a male, and an un-related child female. How this is re-interpreted is not always clear.
Likewise metallurgy can change our view of history. The huntarian museum in Glasgow University has 4 coins, held by past antiquarians and historians to be crude forgeries. They turn out from metallurgy to be real, and futhermore confirm an emperor held for a very long time to be speculative, and undocumented. We now have structural evidence of his existence.
AI tiktoks aren't currently helping here. They aren't providing visuals which carry the history forward, they just project classicly bad tropes. Sure, they're entertaining. So is "gladiator" but historians had enough on their plate there.
This wasn't confusion or mistakes, it was intentionally manipulated for the purpose of creating a more likeable character. Claudius in the series directly contradicts many of the well known traits (and actions) of historical Claudius, such as his rather indulgent excesses in the games and love of bloodshed. But even that's way better than modern BBC 'historical' media where they have a habit of doing things like turning English royalty black.
That's why all of this and then to complain about a short clip having a few mostly irrelevant mistakes (which is not to say the author should not endeavor to correct them) is so utterly transparent as gatekeeping, let alone with such utterly hyperbolic phrasing as calling it "dangerous."
Claudius is played by Derek Jacobi substantively as Robert Graves wrote him. in the 1930s. Graves wrote him as the hero. Of course he was whitewashed. So, what the BBC did is faithfully recreate the characterisation Graves wrote. And wrote, a long long time ago, before historicity was an issue. Graves wrote two books about claudius, he becomes a darker character in the last 10 pages of the second book. Graves also wrote count Belisarius about a different time in a roman story arc. He wrote for an age where many people in England spoke latin, a little, from school. He was writing for the wider audience who maybe didnt but also knew roman values existed in the english cultural landscape, because it was taught.
The danger here, is autodidacts, who take this tiktok as a source of ground truth in their ideation of how the past is. We can see the effect in Vikings-have-horns story. Thats an almost entirely modern construction. It literally has no historic context, horns on helms appear in times, for decorative purposes. We knew from archeology viking helms didn't do this but Hollywood does as hollywood must. Kirk Douglas looked sexy as a viking in a helm with horns and here we are.
Yes, "dangerous" is hyperbole. Everyone arguing a case uses hyperbole, rhetoric was a significant component of the cursus honarum for romans with status to enter the senate. your enemy was full of shit, literally, or more figuratively, because that was how you won. The danger is not the same as the danger of eating poisonous mushrooms, or walking in bare feet on broken glass. its rhetorical danger.
Gatekeeping exists. Objections to it, tend to look very like another rhetorical objection. IF you were permitted into the gatekeeping camp, you'd do it to, and in fact, what you are doing can be said to be a function of using the gate to hold it open to every theory allowed, rather than using the gate to shut off noise. you're therefore gatekeeping too, in a different sense. keeping it open, to suit your own purposes.
What I'm saying is that the BBC intentionally subverts their historical series to intentionally inject biases that are ahistorical but fit either commercial or political tastes. These TikTok videos were made in good faith but made a series of largely inconsequential mistakes inadvertently, and in a ___domain where few expect the sort of accuracy most would expect from something appearing on the BBC.
The latter is far less relevant than the former, yet it's the former pretending to freak out about the 'dangers' posed by the latter. And no I don't gatekeep on anything, ever. It's childish, insecure, and myopic. The more people that get involved in any thing the better that thing will be, in the longrun.
Fair enough. I suppose we won't agree on the relative harms. I tend to focus on bad AI outcomes as indicative of the badness in AI as applied in general. it is actually awesome how the videos are coming out, they are both good, and bad depending on which things you notice. The artifacts of a-historicity are less bad for you, and I think for historians are the niggles which suggest distrust in the product because they project glass windows or pizza into a time they didn't apply to
Awesome to video, and constructed reality. Not so awesome to fine detail. It will be similar for "what would it be like to be inside a nuclear reactor" because it is very unlikely to model the real cherenkov radiation effects, or the decay of the video CCD and your eyeballs, unless suitably prompted and even then, its so subjective, the most likely outcome is the video tropes in millions (hyperbole) that can be scavanged for imagery. Or surgery, which will model what TV doctors do which tries to be faithful, but can only approximate. All the AI models of the world will be nit-picked to death, and subsequently improve.
Of course I agree on all of this. Another typical one people don't realize is when they look at pictures of the cosmos and awe at the wonderful colors, except those colors don't actually exist. If we had a spaceship that could bring us within range to see the Pillars of Creation with the naked eye, they'd probably look mostly blandish gray and perhaps we might pick up a hint of red, and that's that. The colors in most cases are post-processing photoshop efforts done by artists to colorize astronomical data that falls outside the visible spectrum - infrared in particular. Even in cases where it's done in the visible range, it's a high magnification long exposure which again means it'd look nothing like what you'd "really" see.
And yeah, the nitpicking can indeed lead to much better outputs in the future. And also give historians a chance to flex their knowledge of esoterica. And unlike some high production value historical series, it can all be redone with this new input in no time. Of course with new mistakes to then correct, probably endlessly, but there's an asymptote towards perfection we can progress along. When the authors start intentionally breaking things to inject their own politics or biases though, and then presenting that as 'authentic', I'd be the first to be here bitching about it. Well unless the author had some reasonable, if fringe, argument for that being what things were "really" like - history is, in many cases, a recreation based on best guesses, and those guesses end up wrong with a reasonable degree of frequency.
Are you really pushing AI apologism to the point of calling "anality" the concerns of experienced historians? How does this contribute to a reasonable discussion about the dangers of what this video might represent?
I would understand (and not agree) if you said you don't think the videos are dangerous or problematic, but dismissing valid concerns from subject matter experts is just pure tech hybris. And that's quite scary, to be honest, to see how many people already subscribe to this complete lack of critical discussion on the AI video trend.
These videos are viral traps made to make money. They have absolutely no historic base, and yet they are presented as such on social media. I don't care what the authors say, because that's not what 53 million people read when they just see a video that's absolutely inaccurate.
Yeah, and meanwhile BBC are pretty comfortable with using blatant CGI to intentionally mislead people into believing that there are impossibly small cats in the world:
One could critique Dateline or the History Channel by nitpicking accuracy as well. All of those efforts require dedication, and that's all that's standing between a production effort and the homogenization of history.
Yes I do think the dangers of modern day historians are comparable and your comment is a good example of why. You've downplayed professional historians as having "subtle differences" by presumption which highlights the exact danger that I speak of.
The danger of professional historians is that they pretend to not inject bias into their interpretation of history, and a lot of people fall for this via appeal to authority and take their interpretation at face value.
> The danger of professional historians is that they pretend to not inject bias into their interpretation of history
Not true at all. Good professional modern historians are well aware of bias introduced by PoV, material selection, etc. They'll often discuss their own bias and the journals they publish in all too often have Meta Overviews that address the breadth of presentation on particular sibjects.
Here's an example of a profession historian looking at the range of views in the community on the decision to drop atomic weapons:
Funny to choose an example where even the author admits:
> Personally, I don’t wade into these questions much, professionally, or even here on the blog. They honestly don’t interest me very much. Maybe it’s a sign of how post-post-Cold War I am? I don’t know. To me it has always seemed like splitting very fine hairs, trying to make distinctions without much difference. In my mind, the atomic bombings were plainly not ethically very different than the previous firebombings of Japan or Germany
But hey, I’m glad you found an example of a historian that admits there are “hazards” to being a historian. Do you really think this one example of a historian being self aware precludes what I’ve said about historians being dangerous? I think it supports my argument. For every one historian that acknowledges their own bias there are plenty that do not. Just as I would be able to provide examples of amateur historian AI content that even historians would not find problematic. The difference is that I would not use my example to say that AI videos are not dangerous.
Funny in a world of “history” channel where many of these professional historians are used when they can take a break from their regularly programmed conspiracy theories that anybody could with a straight face say they are any less dangerous. The amount of revisionism in history should have been a clue already but how much more defilement of our past can one ignore before calling the entire field out for what it is: a tool for propaganda
How convenient that historians that exhibit behavior that proves my point are not real historians but the ones that exhibit behavior that prove your point qualify.
You seem to believe that professional historians are only so if they support your viewpoint. With that kind of logic how could you possibly be wrong? I didn't realize that we were operating under your magical ability to purity test historians.
Of course it'd be the case that the likes of EP Thompson and Niall Ferguson or Will Durant are passable even with their wildly biased interpretations of history, but dinesh d'souza or david irving are clearly not professionals because, you said so.
You can disagree with expert historians who have researched their field, and sometimes you'll be right.
But an amateur who uses AI and in 4 hours work has produced a complete falsehood seems more dangerous, in that it creates pollution very fast and easily, in a way real (mistaken or not) historians cannot.
It's simply a lot of garbage, made very fast and easily. It has the potential to outcompete everything else.
They are not. It’s quite difficult to prompt a generative model to get details correct, and it takes a ton of tries if you can make it work at all. Practically impossible if the model wasn’t trained with data that makes a distinction in what you’re trying to specify.
So creating accurate generative media is categorically more difficult than inaccurate media.
Because it’s a lot easier to make up shit than to create stuff that is vetted?
Also there are videos made by historians or science educators.
Of the top of my head I can think of Flint Dibble who is an paleontologist? and miniminuteman who is a science communicator who also has an archeology degree?
And while both of them have decent to great success on YouTube, their success pales in comparison to the Ancient Aliens and Atlantis video makers, simply because those are more attractive stories even if they’re not real, but are presented as legitimate theories.
This is a grossly unfair comment. Do you really expect someone who had dedicated their life to an academic pursuit to also be CGI experts and filmmakers?
Why can’t professionals be allowed to criticize without having others fault them?
Edit to add: I’m sure the majority of devs on this site have opinions on current LLMs, is it fair to dismiss everyone’s concerns and tell them “you know how to program, make a better model.”
Consider storytales told around the campfire. Are they true? Before the internet and journalists, you had to decide for yourself.
We are rebounding from a historically brief period of high trust to something more historically normal.
The difference between campfire stories and the world we live in is our world has journalists who fact check and surface falsehoods. Sure, there’s a delay. The work takes a moment. So we hang out while experts do the debunking for us.
As someone technical who works in creative circles, I hear this kind of reasoning all the time and it drives me nuts. The “it’s just art bro, don’t take it so seriously” excuse is so tired, but like the best fake excuses there’s no way to counter it.