If you haven't heard of Edward Fredkin, one of Patrick Gunkel's longest-term friends and staunchest supporters, I'm sorry—but you really should learn a little bit about him.
Fredkin is known for many inventions, both theoretical and applied.
He held numerous patents and—with the help of his wife, Joyce—was a successful businessman.
Among Fredkin's great theoretical contributions, which may yet prove immensely important to humanity, are the concept of "digital physics" and the "Fredkin gate," a reversible circuit configuration that may prove foundational to quantum computing. He also invented the trie, a data structure still in wide use today.
Born in Los Angeles in 1934, Fredkin was a Caltech dropout, computer programmer, and Air Force fighter pilot—just the sort of polymath genius who finds their way to MIT, in other words. He never finished college, yet he still managed to return to academia as a full MIT professor in 1968.
In 1969, Fredkin became a multimillionaire overnight when his company, Information International went public.
Then, from 1971 to 1974, he served as director of MIT’s Project MAC—a Pentagon-funded effort that, with help from AI pioneers like Marvin Minsky and John McCarthy, pushed forward innovations in computer time-sharing and artificial intelligence.
Fredkin kept a long affiliation with MIT even after leaving to work at other universities, including Boston University and Carnegie Mellon University.
Along the way, he founded other companies and bought an island in the Caribbean, where he flew journalist Robert Wright for a 1988 article in The Atlantic Monthly. Fredkin also created a philanthropic foundation and supported a number of his passion projects including, from time to time, Patrick Gunkel.
According to an impressive 2023 remembrance by fellow MIT-connected scientist and businessman Stephen Wolfram, Fredkin may have fallen on hard times later in life.
In 2007, Fredkin sold his Caribbean island to Richard Branson for between $13–20 million. And when Gunkel also fell on hard times in the 2000s, Fredkin reached out to Wolfram in a (failed) bid to help Gunkel get back on his feet.
I never met Fredkin, but he sounds like a charming person who was well-loved and full of colorful stories.
As evidenced by the autobiographical videos he recorded and left behind on YouTube, Fredkin was the sort of person who wanted to create a hero myth about himself and his life.
He wanted to leave a legacy for posterity.
There's nothing wrong with this, in my opinion, especially because of what an interesting life he led and how ultimately important his ideas may turn out to become.
But as much as you may be able to learn about Edward Fredkin and his life, I'm sad to report that it may be a very long time until you're able to learn more about what he did for MIT in the late 1960s and early 1970s—some of the most productive and impactful years of his life, and a crucial period in the history of computing and artificial intelligence.
The reason?
Because MIT’s Distinctive Collections, the archive to which Fredkin donated his MIT papers, has recently locked up his once-public papers and thrown away the key.
What is the MIT Distinctive Collections?
Since late 2019 or early 2020, Fredkin’s papers have been held at the MIT Distinctive Collections archive in Cambridge, Massachusetts.
I’ve been there.
In June of 2024, I hired someone local to Boston to visit the MIT Distinctive Collections and photograph the contents of five folders in Fredkin’s collection.
On the good advice of Professor Betsey Dyer, I had started to think about who Gunkel knew and whether his acquaintances might have preserved some of Gunkel’s material among their papers. Sure enough, it turned out that Fredkin had done exactly this.
According to their website, MIT’s Distinctive Collections:
…collects, preserves, and fosters the use of unique and rare materials… to cultivate the interest in past, present, and future.
At some point a researcher like me orders specific folders for research.
An appointment is made.
The researcher then shows up to the Distinctive Collections reading room and gets served the documents.
Although MIT has fairly strict rules about the Distinctive Collections materials—you have to put any bags or backpacks in a locker, you can’t assume that everything you see is free to copy and distribute, etc.—it overall seems fair and sensible.
Unfortunately, as it turned out, two of the five folders associated with Gunkel were labeled “restricted access” by the Distinctive Collections at the time I hired someone to photograph them. The remaining three folders were public.
For the restricted access folders, an archivist informed me they were restricted for “personnel” reasons and would be available to the public 75 years after their creation date—in 2053, in other words.
Although I was disappointed about the restricted folders, they didn’t bother me a whole lot.
It’s difficult to be sure, but I felt “personnel” reasons probably related to Gunkel’s hiring on Project MAC in 1971 and his subsequent firing in 1974.
It’s very likely that, at some point, there was an incident related to Gunkel and the police were called.
In my opinion, MIT would be well within its rights to restrict human resources-related files like these because they would pertain to the operations and reputation of the university rather than any substantive ideas or work that the public would be interested in knowing.
Especially with respect to Project MAC and its Pentagon ties, some level of vetting or background research may also have been appropriate, and there’s probably a decent reason to keep this sort of record confidential for an extended period of time.
From “Partial Access” to “No Access”
Ever since this foray into Fredkin’s files in 2024, I was pretty sure I’d gathered all the Gunkel materials available at MIT.
But last week, for some reason, I decided to jump back on the Distinctive Collections website again to make sure.
Not in my wildest dreams did I expect to find less accessibility than before, but that’s exactly what I found.
With a shock, I realized the entire collection of Fredkin’s papers—4 boxes, 4 cubic feet, perhaps 200 folders worth of material from the early 1970s—had been completely restricted:
The only clue was a note that in April of 2026, a “formal review” had occurred and as a result the entire collection was put onto lockdown.
What’s even crazier is that the finding aid had also been removed, so not only was the collection inaccessible, but you couldn’t even know what was in it anymore.
To try to clear things up, I paid a visit to the gloriously important Internet Archive.
Here, I was excited to find that some of the pages from the Fredkin collection as I found them prior to April 2026 had been preserved as digital captures.
Based on my search, it looks like the collection first became available in early 2020—over three years before Fredkin’s death in June 2023.
And up until April 2026, the access status had been “open” unless otherwise restricted, as this screenshot from August 2024 shows:
Most remarkably, the Internet Archive reveals that in March 2023, a Distinctive Collections archivist actually went through the Fredkin papers and made updates.
The folders remained, on the whole, public.
In fact, the archivist may have even updated the finding aid to make it even easier to know what was in the files.
At first I wondered whether MIT had originally made a mistake in providing public access to the materials. This will happen from time to time, especially with library resources stretched thin.
But what I found on the Internet Archive confirmed the initial public availability of the folders was no mistake: the Fredkin collection didn’t get accidentally posted online, it was actively vetted by the Distinctive Collections staff and a decision was made that it could continue to remain public.
And that’s how things sat for a few more years until, for no clear reason, the gate slammed down and access to everything was yanked in April.
“Changing Uses of Our Collections…”
I was starting to get pretty irritated, even angry.
Conspiracy theories about powerful people, valuable patents, and embarrassing blackmail swirled through my head.
But after some back and forth with librarians at the MIT Distinctive Collections, on Friday I received a fascinating statement from Mikki Macdonald, the Institute Archivist, which I am going to share in full because I believe this signifies a sort of breaking news that We the People of the Transformer Age may need to sit up and pay attention to:
Mrs. Macdonald wrote:
Changing uses of our collections in the time since many were originally donated have prompted a review of our access policies and the way they are applied across our collections. In the course of this review, we made the decision to close the Fredkin collection until we have confirmed with the family about their access preferences for the materials.
Fredkin donated his collection in 2019.
He died in 2023.
He made the decision to donate his papers personally while he was alive; his papers remained accessible while he was alive; his family, presumably, didn’t have anything to do with it.
What sort of “changing uses” for the Distinctive Collections could have possibly happened between now and 2020, a span of only six years, that would have resulted in a need to consult with the family?
Oh yeah… right.
And no, I’m not referring to the coronavirus pandemic.
I’m referring to the invention and growth of large language models (LLMs), and their ravenous requirements for original data and information.
Given how shady the Frontier AI companies have already proved themselves in acquiring the information they use to build their models, it’s totally reasonable to imagine them sending out people to covertly to systematically photograph entire archival collections.
I’m just guessing, here, by the way.
I don’t know if any specific incident happened at MIT’s Distinctive Collections to prompt their April 2026 “formal review,” but I can imagine someone coming back, day after day, making scans of every single page in Fredkin’s collection and refusing to state what, exactly, their research purpose was.
And I know that Leo S. Lo, Dean of Libraries and University Librarian at the University of Virginia, wrote a paper in April 2026—the same month the Distinctive Collections conducted its formal review—entitled “Memory Without Origin: Provenance, Consent, and Trust in the Age of Generative AI.”
In the paper, Lo referenced the fact that AI companies are approaching large libraries offering to fund digitization in exchange for training data and identified a number of potential risks, including the unusual concern that information becomes divorced from its source, resulting in a situation where models produce knowledge without any understanding of where it came from.
“Current AI foundation model training permanently eliminates an institution’s ability to govern how its collections are used,” Lo wrote, “unless contractual protections are established in advance. It is a stewardship, legal, and ethical decision, not an access decision, and it cannot be undone.”
The Implications Are Astonishing
I started out prepared to be angry about this, but to be honest—assuming, again, that I understand what’s happening—I believe MIT Distinctive Collections is absolutely doing the right thing here.
Access is an important part of their mission statement, but treating their unique historical materials like so much wood pulp to be fed into a machine that will smooth it all out, manufacture it, and process it into a glistening white sheet upon which anything can be written is probably not what anyone who donated to the Collections ever had in mind.
If MIT actually does reach out to Fredkin’s family, I assume it’s because Fredkin did not realize what these companies would be doing just a few years after he donated his papers, and the archivists need to confirm what level of permission Fredkin would have wanted given the increasing value Frontier AI companies are placing on information their models have not yet been ingested.
Unfortunately, the rapidly moving field of AI appears to be butting up against a probably slower-moving field of archivists and librarians, many of whom are probably far more hard-pressed to resist the cash-for-content offer being made by these wealthy companies.
This could be one of those things that just gets quietly absorbed into society without much of a fuss, “the way things are.” Suddenly we look up and there are far more digital resources than existed before, AI models are a lot more robust, and some have even created licensing deals that allow them to claim exclusive ownership over certain unique data sets.
On the other hand, as the concept of “digital immortality” continues to gain prominence, distinguished people may become afraid of donating their papers because of the risk that it gets used in a way they can’t foresee.
The scenario that may end up playing out has been called enclosure.
Historically, when a resource that was once accessible becomes valuable, holders fence it off.
If archivists, families, and estates come to see "openly available" as synonymous with "will be ingested, decontextualized, and possibly used to create digital versions of dead people," restriction will inevitably follow.
This means that the already fragmented field of research may become even more fragmented, with varying tiers of quality and a persistent confusion about what constitutes legitimate, traceable knowledge.
Some academic fields may feel this acutely.
Contemporary history, biography, sociology of recent events, and anything relying on personal papers of living or recently deceased people could see a "20th-and-21st-century gap" where the most important sources are the hardest to find.
This is already a problem, of course, but if anything it will most likely become exacerbated.
Imagine the irony—it’s 2040 A.D., and the recent past, despite all our electronic storage options and methods of self-expression, has become much more difficult to study and understand than the world of 200 years ago.