No law is going to force criminals or those spreading political propaganda to label their output as “AI-Generated”. And yet this is the most destructive misuse of AI.
This seems like a very real problem. We are all trying to learn how to not accept material as "real" until we verify whether it's AI generated or not-- but it wouldn't have crossed my mind to direct that skepticism at an "old" photo. Having looked at your examples, I see that we have to be skeptical across the board now, which is kind of sad.
Kudos for fighting for your craft. You’re absolutely correct that accurately documenting history supports critical thinking, liberty and respect for the past that paved a pathway for us📷
Marina, I don’t think you’re being cranky to call out this very disturbing use of AI and how it will impact what people think is historically accurate. History is indeed a contested narrative space, but the difficulty in figuring out what has been changed in a AI fake photo has the potential to shift what is remembered and archived.
I like your self-aware discussion of the pushback you initially received regarding colorizing old photos. The questions raised to you were valid, but my response is that colorizing amounts to the way history is dramatized in prose accounts, be they narrative nonfiction or novels - as a viewer and reader, I like the feeling of being in the past, as long as the author tells me what has been changed for dramatic effect and it’s possible to fact check.
But as you write here: “AI-generated photos, on the other hand, have the potential to create entirely new realities out of nothing. They're not just tweaking or reinterpreting existing historical records - they're fabricating them from scratch. And when those fabrications are convincing enough, they can be incredibly difficult to detect or debunk.”
P.S. DM me if you’re interested in being interviewed for a project about AI that I’m working on :-)
will soon lead to fake news video, film clips altering history to ones liking....the ongoing fight between nature and man on a sometimes incredibly beautiful planet...
There are legitimate concerns, but I think we also need to take a step back and re-evaluate. What is the difference between AI fakes and Photoshop fakes? Nothing, really. AI can just make them quicker. Photoshop didn't bring about some catastrophe and I don't think AI fakes will either.
As the author says in the article: "Thankfully, as time went on and the technique became more refined and understood, the initial shock gave way to appreciation and curiosity. And it's been really cool to see this happening."
I think we're likely to experience the same with AI.
"But here's the thing: while traditional forms of historical distortion, like propaganda or revisionism, can certainly be influential, they still operate within certain bounds of plausibility" I completely disagree. One needs merely to look at the rhetoric on Jan 6 riots and see how completely out of bounds of plausibility some rhetoric is even with all the photo and video evidence widely available.
This is going to be highly problematic for historians, and historical researchers. In particular, my primary concern is what unscrupulous and unethical “researchers” will do with the technology. On one hand, it may lead to falsifying research to complete a book about a historical event or person. For example, my recent book I was unable to locate photos of the main subjects (but I have evidence that they existed at one point). An unethical researcher and writer could use AI to generate realistic looking images, especially given that there are drawings of the people. They could pass them off as authentic - an amazing discovery.
More chilling is what the historical deniers could do with the technology. Image the damage Holocaust deniers can, and no doubt will, do with the technology. They will be able to use alleged photographic evidence to support their agenda or refute the true historical record.
Most consumers of information don’t fact check or verify the authenticity of information or an image. It is going to get to the point where the real versus AI become so intermingled that true historians will find themselves a minority fighting to keep the historical record as authentic and verifiable as possible.
History is interpreted, and often revised as new information is discovered or different perspectives examined. Historical figures can be vilified by some researchers and praised by others. The interpretation of historical facts is key. And if we can’t rely upon the alleged facts (in this case photographs) then we can’t present at least a reasonable interpretation of certain events.
And what about when documents are AI generated to look like they were produced during a specific time period and replicating signatures, etc.?
AI is going to have a big impact on the historical community. And it isn’t going to be a good one.
And my view of colorization is that it makes old images accessible and relatable to a large audience. We connect to the people when we see them in color - to see them as they would have looked - rather than BW. My oldest sons are almost 21… they feel bad for me that I had a BW TV in my room as a kid! BW doesn’t have the same
Impact on their generation as does color. As a researcher, I prefer the originals to use in my own work. But a professionally colorized version is an excellent adjunct. I’ve paid to have a few colorized and it changes the connection I have with the person in the photo.
My question is, if these AI-generated historical photographs are out there on the internet without notification that they are AI-generated, are the next generation of AI tools being trained on them? I mean, are they part of the content that is used for training AI tools?
I appreciate your thoughts on this. I try to question erroneous or flat out false information in text format when I come across it but it will be beyond my ability and resources to do the same for every photograph I see on the internet. Images will become too good to provide clues to badly generated fakes. How would I even know whether the sources I use to verify the image are using the doctored image or the original? I don’t like the idea of censorship but I believe the integrity of words and images must be protected from blatant deceit.
Your description of people reacting to your colorization of photos is wild, as colorization has been around since photography was invented. I remember having a bit of a problem when old black and white movies were being colorized by Turner back in the day, but most people who complained about that were arguing that b&w movies has their own aesthetic and feeling, which was ruined by colorization (and yes, in spite of people like Méliès colorizing their movies as far back as 1897). AI creating reality from scratch is much more frightening and disturbing, and I'm not sure how we can stop it.
Agreed. While exploring what's possible in the AI ecosystem is an insightful beta-test, as you've proven here, there needs to be a limiting agent on the extent to which AI algorithmic manipulation interferes with and/or replaces altogether historical archives. It's yet another component to consider in restoring public trust in media.
... that this power of A.I. to churn out fakes, will heal us from all the news feeds and content spread through the social media, and will reinforce the professional and trustworthy sources of information.
If anything of unknown provenance may be a fake, there is no point reading the feeds at all. Renown newspapers and magazines do cross-check information, and either publish it as true, putting their reputation at stake, or mention that the source or credibility of an image could not be verified (see the BBC, or The Guardian).
Will we finally turn our backs on all the facebook-like news feeds, and come back to papers and magazines (they may be online), even if we must pay for them, like we used to pay for every paper bought at a newsstand?
Yeah it'll become a huge problem and it already is to an extent.
I've seen it mainly on Facebook where I get astronomy accounts that show amazing images of planets and moons, without telling the audience that it's artistic impressions.
I've also seen it on Youtube where a react channel reacted to some "James Webb telescope images", which looked amazing. The problem was that at the time, JWST had only just begun to do a few test images, so it would've been impossible. Me and a few others tried to argue this in the comment section, but were largely ignored and the "wow this looks amazing"-comments got far more likes..
I'm glad that Youtube is starting to address this AI business where people have to declare when they post AI-altered footage or images or risk going down a road to getting their account banned - at least when the content is for educational purposes and not just entertainment.
Even before this, I've seen such educational channels mark artistic images with a "CGI" in the corner, which helps a lot.
It's both extremely awesome to have AI help toy around with reality and make long-dead things come alive, but it's an obvious double-edged sword that is scary as heck, because so many seem to be ready to use it in a bad way, like criminals and dictators.
It's interesting that for all the efforts of various organizations like OpenAI to pursue what they consider "safety" they've never made any serious attempt to instill the concept of Truth - let alone Historical Truth - into the AIs. But how would you?
No law is going to force criminals or those spreading political propaganda to label their output as “AI-Generated”. And yet this is the most destructive misuse of AI.
This seems like a very real problem. We are all trying to learn how to not accept material as "real" until we verify whether it's AI generated or not-- but it wouldn't have crossed my mind to direct that skepticism at an "old" photo. Having looked at your examples, I see that we have to be skeptical across the board now, which is kind of sad.
Kudos for fighting for your craft. You’re absolutely correct that accurately documenting history supports critical thinking, liberty and respect for the past that paved a pathway for us📷
Marina, I don’t think you’re being cranky to call out this very disturbing use of AI and how it will impact what people think is historically accurate. History is indeed a contested narrative space, but the difficulty in figuring out what has been changed in a AI fake photo has the potential to shift what is remembered and archived.
I like your self-aware discussion of the pushback you initially received regarding colorizing old photos. The questions raised to you were valid, but my response is that colorizing amounts to the way history is dramatized in prose accounts, be they narrative nonfiction or novels - as a viewer and reader, I like the feeling of being in the past, as long as the author tells me what has been changed for dramatic effect and it’s possible to fact check.
But as you write here: “AI-generated photos, on the other hand, have the potential to create entirely new realities out of nothing. They're not just tweaking or reinterpreting existing historical records - they're fabricating them from scratch. And when those fabrications are convincing enough, they can be incredibly difficult to detect or debunk.”
P.S. DM me if you’re interested in being interviewed for a project about AI that I’m working on :-)
will soon lead to fake news video, film clips altering history to ones liking....the ongoing fight between nature and man on a sometimes incredibly beautiful planet...
There are legitimate concerns, but I think we also need to take a step back and re-evaluate. What is the difference between AI fakes and Photoshop fakes? Nothing, really. AI can just make them quicker. Photoshop didn't bring about some catastrophe and I don't think AI fakes will either.
As the author says in the article: "Thankfully, as time went on and the technique became more refined and understood, the initial shock gave way to appreciation and curiosity. And it's been really cool to see this happening."
I think we're likely to experience the same with AI.
"But here's the thing: while traditional forms of historical distortion, like propaganda or revisionism, can certainly be influential, they still operate within certain bounds of plausibility" I completely disagree. One needs merely to look at the rhetoric on Jan 6 riots and see how completely out of bounds of plausibility some rhetoric is even with all the photo and video evidence widely available.
This is going to be highly problematic for historians, and historical researchers. In particular, my primary concern is what unscrupulous and unethical “researchers” will do with the technology. On one hand, it may lead to falsifying research to complete a book about a historical event or person. For example, my recent book I was unable to locate photos of the main subjects (but I have evidence that they existed at one point). An unethical researcher and writer could use AI to generate realistic looking images, especially given that there are drawings of the people. They could pass them off as authentic - an amazing discovery.
More chilling is what the historical deniers could do with the technology. Image the damage Holocaust deniers can, and no doubt will, do with the technology. They will be able to use alleged photographic evidence to support their agenda or refute the true historical record.
Most consumers of information don’t fact check or verify the authenticity of information or an image. It is going to get to the point where the real versus AI become so intermingled that true historians will find themselves a minority fighting to keep the historical record as authentic and verifiable as possible.
History is interpreted, and often revised as new information is discovered or different perspectives examined. Historical figures can be vilified by some researchers and praised by others. The interpretation of historical facts is key. And if we can’t rely upon the alleged facts (in this case photographs) then we can’t present at least a reasonable interpretation of certain events.
And what about when documents are AI generated to look like they were produced during a specific time period and replicating signatures, etc.?
AI is going to have a big impact on the historical community. And it isn’t going to be a good one.
And my view of colorization is that it makes old images accessible and relatable to a large audience. We connect to the people when we see them in color - to see them as they would have looked - rather than BW. My oldest sons are almost 21… they feel bad for me that I had a BW TV in my room as a kid! BW doesn’t have the same
Impact on their generation as does color. As a researcher, I prefer the originals to use in my own work. But a professionally colorized version is an excellent adjunct. I’ve paid to have a few colorized and it changes the connection I have with the person in the photo.
My question is, if these AI-generated historical photographs are out there on the internet without notification that they are AI-generated, are the next generation of AI tools being trained on them? I mean, are they part of the content that is used for training AI tools?
I appreciate your thoughts on this. I try to question erroneous or flat out false information in text format when I come across it but it will be beyond my ability and resources to do the same for every photograph I see on the internet. Images will become too good to provide clues to badly generated fakes. How would I even know whether the sources I use to verify the image are using the doctored image or the original? I don’t like the idea of censorship but I believe the integrity of words and images must be protected from blatant deceit.
Your description of people reacting to your colorization of photos is wild, as colorization has been around since photography was invented. I remember having a bit of a problem when old black and white movies were being colorized by Turner back in the day, but most people who complained about that were arguing that b&w movies has their own aesthetic and feeling, which was ruined by colorization (and yes, in spite of people like Méliès colorizing their movies as far back as 1897). AI creating reality from scratch is much more frightening and disturbing, and I'm not sure how we can stop it.
Very interesting. Authentication of actual historical records via blockchain may help (https://blockchain-observatory.ec.europa.eu/document/download/c63a0010-6080-41bc-af26-4d8f9382c629_en?filename=EUBOF3.0_Decentralized%20social%20media_Final.pdf), but unclear whether this will be widely adopted. Ideally one should get to the point that if it does not have an original bona-fide blockchain authentication, it can be assumed to be false.
Agreed. While exploring what's possible in the AI ecosystem is an insightful beta-test, as you've proven here, there needs to be a limiting agent on the extent to which AI algorithmic manipulation interferes with and/or replaces altogether historical archives. It's yet another component to consider in restoring public trust in media.
I have a dream...
... that this power of A.I. to churn out fakes, will heal us from all the news feeds and content spread through the social media, and will reinforce the professional and trustworthy sources of information.
If anything of unknown provenance may be a fake, there is no point reading the feeds at all. Renown newspapers and magazines do cross-check information, and either publish it as true, putting their reputation at stake, or mention that the source or credibility of an image could not be verified (see the BBC, or The Guardian).
Will we finally turn our backs on all the facebook-like news feeds, and come back to papers and magazines (they may be online), even if we must pay for them, like we used to pay for every paper bought at a newsstand?
Fake news are everywere. I think we must turn back education to librarys ann books, instead of screens
Yeah it'll become a huge problem and it already is to an extent.
I've seen it mainly on Facebook where I get astronomy accounts that show amazing images of planets and moons, without telling the audience that it's artistic impressions.
I've also seen it on Youtube where a react channel reacted to some "James Webb telescope images", which looked amazing. The problem was that at the time, JWST had only just begun to do a few test images, so it would've been impossible. Me and a few others tried to argue this in the comment section, but were largely ignored and the "wow this looks amazing"-comments got far more likes..
I'm glad that Youtube is starting to address this AI business where people have to declare when they post AI-altered footage or images or risk going down a road to getting their account banned - at least when the content is for educational purposes and not just entertainment.
Even before this, I've seen such educational channels mark artistic images with a "CGI" in the corner, which helps a lot.
It's both extremely awesome to have AI help toy around with reality and make long-dead things come alive, but it's an obvious double-edged sword that is scary as heck, because so many seem to be ready to use it in a bad way, like criminals and dictators.
It's interesting that for all the efforts of various organizations like OpenAI to pursue what they consider "safety" they've never made any serious attempt to instill the concept of Truth - let alone Historical Truth - into the AIs. But how would you?