X

NFTs and AI are unsettling the very concept of history

AS AN ARCHIVIST, I’m excited about what disruptive innovations like non-fungible tokens (NFTs) and artificial intelligence may mean for archives. But I’m also worried. These developments pose existential threats to our field, and by extension, to the survival of human history and culture.

give old films away for free. It started in 1999 when I was seduced by the promise, excitement, and just-felt-rightness of the gift economy. Not 30 seconds after we first met, Internet Archive founder Brewster Kahle asked me, “Want to put your film archives online for free?” Braving the new world of video digitizing and sputtery streaming changed my life. Our archival footage enabled thousands, maybe millions of artists, videomakers, educators, and even post-Communist Polish village kids to remix history and bring the past into the present. I never knew how many people were using our material or who they were—but wasn’t that the point?

In 1999 the future of our archives was to be consumed, to enrich public memory with new evidence without hassle. I wanted our archives to be as ubiquitous as infrastructure, to work their way into every corner of the net, to propagate everywhere without need for attribution or credit. I wanted our archives to vanish in the web.

I still do.

But now the survival of archives as we know them is uncertain. Whether we know it or not, we all rely on a patchwork of chronically underfunded public and private institutions that hold the world’s histories and cultural heritages in trust for all of us and make them accessible. Every time we see an old photo, hear a historical recording, see a news clip, or find a family history document, it likely originated in an archive. While we see and touch massive digital archives online, most archives are still largely undigitized collections of physical media like film, video, music, photographs, and paper documents. By design, archives are deliberate and thoughtful, with a timeline designed to preserve culture “forever.” They’re not built to nimbly weather disruption.

It was only a matter of time before the market figured out a way to manufacture and sell digital scarcity, and the marketplace for cultural objects has moved well past the archival ecosystem. Artists, gamers, entertainers, athletes, and executives now sell NFTs, tokenized digital objects whose authenticity is said to be assured by the reverse traceability of blockchain transactions. The combination of Covid-19 isolation and cryptocurrency profits created a powerful incentive for digital-positive collectors to compete for these NFTs, and some creators are raking in Ethereum.

Law professor Tonya M. Evans optimistically suggests that crypto art offers Black artists and communities opportunities to bypass white art gatekeepers and “capture and own the value of the culture that they produce.” While the current boom may well go the way of the 1920s Florida land-rush hype, NFTs are the first step in what’s likely to be a robust market for unique or scarce digital objects. Many of these digital objects won’t be born-digital; instead they’ll be one-off digitized copies of physical materials, for which there could be a huge market. Who wouldn’t want to own the master digital copy of their favorite author’s journal, a photograph of Abraham Lincoln or Frederick Douglass, or the recently rediscovered newsreel of the 1919 Black Sox scandal?

Nothing could be a greater cultural and ethical shock to archives than NFTs. Prevailing archival ethics generally dictate that all users are treated equally, and that archival materials aren’t exposed or sold only to high bidders. And once archives select materials for retention, they consider themselves in most cases ethically bound to do so permanently.

If an archive has a merch business, it’s tiny: keychains and postcards. As poor a fit with archival DNA as tokenizing archive collections as NFTs may be, the possibility of leveraging digital scarcity by selling NFTs while retaining physical materials is a hefty temptation. The archival world is a world of inadequate budgets and financial constraint, filled with underpaid workers and massive, poorly resourced projects like digital preservation, and the challenging task of digitizing analog materials. Will archives be tempted by the potential upside of NFTs and tokenize digital representations of their crown jewels (or the rights to these assets)? This would worsen an already bad situation, where institutions like our Library of Congress hold physical copies of millions of films, TV programs, and recordings that can’t be touched because someone else holds the copyright. Ideally, archives and museums should own and control both the physical and digital states of its collections. That won’t happen if they have to sell or license NFTs in order to survive. And there’s another risk: Minting NFTs requires a lot of energy (though we may hope for a cleaner process), and the future security of archives is threatened by climate change. Researchers have discovered that almost all archives will be affected by risk factors like sea level rise, increased temperature, or heavy rainfall.

For those working with the raw materials of history, integrity and authenticity are the chief necessities. How do NFTs address these? While the blockchain is supposed to draw an unbroken link between creator/tokenizer and purchaser, it’s just a record of transactions that might be tainted or even bogus. We know the original Mona Lisa resides in the Louvre, but it’s very hard to identify who really created and who owns many of the millions of creative works made in the analog era. It’s nearly impossible to track who owns the even greater number of digital works made every year. So while the blockchain can help us follow sales and transfers of digital objects, how do we know the original representations made about these objects can be trusted? Already there are many NFTs on offer that are nothing more than tokenized versions of works belonging to third parties, often scraped from museum websites. This isn’t a new problem—after we put digitized films from our archives online for free, many other stock footage companies downloaded them and sold them as their own. Who will be the arbiter of which copies are closest to the original? The quagmire recalls Philip K. Dick’s The Man in the High Castle, where American craftspeople build near-perfect counterfeits of American cultural collectibles. Authenticity is based on aura, aura on belief. If many Alices sell tokenized, pirated copies of the Zapruder film to many Bobs, this bombs the blockchain with many transactions that obfuscate the fact that these might be pirated versions. The solution, obviously, is to know your source—to authenticate objects and their provenance by authenticating their owners. But can that scale in a marketplace where transactions might number in the billions?

Registries are emerging to authenticate sources and provenance, and perhaps even indemnify purchasers against false representations by sellers. These have long existed in the collectibles business. Rare coins are frequently processed by trusted grading and authentication services, which charge to inspect coins and then encapsulate them in sealed plastic slabs. While I can see this happening for the four existing color versions of Edvard Munch’s The Scream, how would it work for millions of new digital works? Will these registries be proprietary or open, and how much will their services cost? And authentication systems ultimately rest on the accuracy of information they receive. Who would arbitrate conflicting claims between potentially millions of squatting bots feuding over provenance and authenticity issues? And could such registries be flooded with blockchain-authenticated look-alikes and deepfakes? If you think this is improbable, just look at YouTube, where Content ID, its suite of pattern-matching algorithms that’s claimed 800 million videos since its launch, supposedly enforces the rights of copyright owners by flagging unauthorized uploads. The system generates huge numbers of false positives and won’t authorize legitimate fair uses of content. And copyright to millions of videos (including many public domain works) is claimed by squatters. Fuzziness and a high error rate may be OK for a commercial service like YouTube. But it’s a disaster for cultural and historical institutions that derive income from reuse. How would a blockchain full of false transactions complicate ownership and authenticity?

One working solution is for cultural and historical institutions like archives to run their own trusted registries of digital objects. But this is expensive, and it creates further incentives for archives to monetize their holdings and become less accessible to noncommercial users, like genealogists, the group that uses archives more than anyone else. If the need to administer NFTs makes NFTs inevitable, that’s a loss. Today’s archivists are faced with the need to come to terms with NFTs without having the resources to ensure their needs are met.

NFTs make archival authenticity, and thus history, more fragile. And right now authenticity is threatened by AI.

If you’ve seen Denis Shirayev’s upscaled historical videos, you’ve seen the past enhanced by a touch of the future. He takes videos scanned from very old films, like our poignant A Trip Down Market Street Before the Fire, shot just days before the 1906 quake and fire that devastated San Francisco, upscales them to 4K, smoothes out jitter and adds color. (Today any video editor can make something almost as good using off-the-shelf tools like Topaz Video Enhance AI.) Shirayev’s videos are beautiful and compelling, but they show you something that never was. They’re not archival; they’re fiction. Artist and author Gwen C. Katz recently demonstrated the flaws of AI colorization, showing that software substitutes a drab color palette for the brilliant colors of historical reality, and pointing out that only primary historical evidence unaffected by our preconceptions can present a plausible image of the past. Will prettified AI-enhanced digital objects made to draw in the eyeballs of distracted web surfers push the original, less attractive evidence out of view? Will people modify archival materials in such a way as to marginalize their original source? Will established archives sit by as others embellish their collections into prettier, more marketable objects? And, worst of all, will pretty images stand in for the inconvenient, hard-to-view authentic record?

The question of “real” vs. “pretty” recently exploded when a Vice article (now removed) revealed Matt Loughrey‘s colorization and alteration of mug shots taken in a Khmer Rouge torture center just before their subjects were murdered. Cambodian genocide survivors led a protest against Loughrey’s project, which added smiles to a few victims’ faces, and the Tuol Sleng Genocide Museum and the Government of Cambodia called for his apology and for the photographs to be taken down. From the viewpoint of historical and archival authenticity, Loughrey’s call for infusing archival photos with a sense of contemporary realism was perhaps most disturbing. “The image that we’ve come to accept as standard is becoming obsolete owing to the advance in display technology,” he told the Daily Mail. “When we consider a museum or a library or a documentary, as these displays advance, which they are rapidly, the producers of these are going to be less inclined to display and use these images. They’re going to have to find new images by repurposing them.” The actual archival record, in other words, can no longer stand by itself.

Together these two developments pose an existential threat to archives. Archives won’t go under, but it’s going to take serious thinking, significant new funding, and public education to help these under-resourced, deliberate organizations respond to rapid change. The integrity and survival of these important institutions are in everyone’s interest. Archives aren’t perfect, but they can help keep us honest. They’re a force for historical accuracy and accountability, if we trust the records in their collections and know their provenance. All of us should hope the promise of NFTs and AI won’t slide us into a world of “fake archives” and speculation in archival collectibles.

I still want our film archives to be consumed freely by makers in familiar media and in media we haven’t yet begun to imagine. I still want our archives to vanish in the web. But I don’t want history and the institutions that painstakingly preserve it to disappear into an eternal and amnesic present.

Rick Prelinger