This is really exciting. If you're not aware, the scrolls at Herculaneum are an entire pagan library from the first century. They're burned, hard to recover, mostly still buried. Being able to decode them without physically digging them up and damaging them is awesome.
Who knows what we could find. So many books have been lost.
Seems a bit confusing to call it a "pagan library" when it's just the personal library of a very rich ancient Roman.
The ancient Roman elite often had extensive personal libraries which they shared with their friends, almost like a very primitive book publishing industry.
From the Christian, anyone who wasn't part of the Christian or Jewish faiths was considered pagan. In the first century CE, Epicureans were part of the broader category of Hellenistic pagan philosophers—which included Stoics, Platonists, and others—who were polytheistic or at least non-Christian. So since Philodemus makes up most of the library here, it's pretty safe to call it a "pagan" library.
I genuinely don't at all understand what the issue is with calling it a pagan library. It's a clear and substantive descriptor.
The history of the Roman Empire can be divided into two overarching periods - the pagan era and the Christian era. A pagan library is significantly more valuable than a Christian library, as we have far fewer pagan texts than Christian ones, and the few we do have were often corrupted by later Christian copyists. The word pagan is doing meaningful work in the sentence.
I actually find this strange tendency of online commenters to link something to their own obscure interest very amusing. It's been a classic for as long as I recall but I encountered another today which I thought was very entertaining where a commenter remarked that he only just realized that "Suno" is the Hindi word for what we'd say in Latin as "Audi". In Latin! Hahaha!
I have decided that I, too, shall use obscure things as benchmarks and references. It's pretty good fun. In this post-Ragnarok-Online world one can imagine we need more such milestones to judge other things by.
In the sense that the later Christian Romans were very very eager to lean on the good ol' book burning.
One of the reasons only a fraction of a percent of the classical texts reached our days is the fact that Christians suppressed those texts, directly (by destroying them) and indirectly (by closing the libraries and temples and institutions of learning which preserved those texts).
In context its clear what is meant and that terminology has a long history.
Sure you could argue the terminology is very christian-centric, perhaps even offensive to pre-christian romans, but quite frankly that's a very uninteresting debate compared to the topic at hand.
> Being able to decode them without physically digging them up and damaging them is awesome.
Can you provide some citations on the technology being used in situ without digging up? As far as I understood this is the application of technology widely popularized by the Herculaneum Challenge, where scrolls are still physically dug up, and x-rayed (which will slowly still damage the scrolls) but without physically breaking them open as was repeatedly attempted in the past.
I don't care much about the slow damage from x-rays: as long as the content is succesfully extracted, one can imagine little other use for the scrolls as is.
I mostly hope some lost works on mathematics will be recovered..
They have been dug up already. IIRC they have undergone extensive scanning over the years - X-Ray, CT etc - while not being unrolled.
So they have the scans of the rolled up scrolls, this is "just" (ha!) using the scan data with lots of algorithms and compute (AI? I presume so) to virtually unroll the scrolls and read the ink off the page.
Actually, they are CT scanning them as the project continues. IIRC they reported about scanning a new (big) batch of them about a month ago.
You are right about not unrolling them though. Many scrolls were destroyed in previous attempts to unroll them physically, so it is fascinating to see how the technology has progressed to allow reading without unrolling.
There is a PBS documentary about this very thing and how it got started. Very cool and worth the watch. Needless to say, the researcher had quite a few hurdles to overcome.
Researchers are further refining the image using a new segmentation approach in the hopes that it will improve the coherence and clarity of the lines of text currently visible, and perhaps reach the end of the papyrus, the innermost part of the carbonised scroll, where the colophon with the title of the work may be preserved.
So the new article is indicating they were able now to decipher the title, and also indicates maybe why the title was not the first thing deciphered (presumably it is hardest to read the innermost parts.)
I'm curious why the title is in the inside of the scroll. That implies you have to completely open it to read the title - is that the way scrolls are usually written?
If you're rolling the scroll up as you read (or, I guess, write) it, you'll leave it rolled up like that. I presume you're expected to rewind the tape before you return it, if you borrowed it from somebody else.
Maybe the title was on both ends. That way, you'd still be able to tell what it was quickly even if not rewound. And the one on the outside just burned away.
> Both images were independently reviewed by the Vesuvius Challenge papyrological team, led by Federica Nicolardi. The simultaneous reproduction of the title image from multiple sources, along with independent scholarly review, provides a high degree of confidence in the reading.
You have 2 teams using the same data, getting to the same conclusion. You also have an author that's known from other sources, with writings that we already "have". Then you have a team of experts reviewing this. Chances are these are real findings and not "hallucinations". Not everything in ML is gen-ai...
To add to this, the main ML parts, as I understand it, are for the initial unrolling of layers, and for the detection of ink vs. no-ink (the position of the 'splotches'). Both of these are trained and calibrated from human observations.
All interpretation of ink as Greek letters is done purely by human inference. This may lead to errors, especially in parts where the ink is preserved especially poorly or where the text is totally different from expectations, but it would be classic human error instead of AI hallucination.
>You also have an author that's known from other sources, with writings that we already "have".
Well, that's exactly what you'd expect from a hallucination no? If the model is overfit enough on the relevant corpus, a title that already exists should be much more likely.
It is not very common to find pre-Kanishka works. I hope that we get some insight into human lives around this time. One of the things I find fascinating about ancient times is how similar humans of then were to us. Akrotiri (similarly preserved by volcanic eruption) was millennia before even the works in this discovery and yet seemed strangely familiar and normal when visiting.
Brent Seales was my second CS professor and taught me how to do OOP in C++. It’s always cool to randomly see the work he’s done every few months. He was working on this project nearly twenty years ago.
> Using ‘virtual unwrapping’, the scroll PHerc. 172 which is housed at the Bodleian Libraries at the University of Oxford has been identified as On Vices by the Greek philosopher Philodemus
As a history nerd and jaded software developer, I've been wondering a lot lately how I can use my tech skills for archeological research. Is there any way for someone with most of a bachelors to get into this kind of thing?
My experience with academia is that most of this hard work is done by undergrads, and conception and management by professors; developers aren't hired to do this. So besides "going back to school", there's no way in for an outsider.
They say they're hiring but I didn't even get an email back about my application and I've been awarded $20k through the vesuvius challenge and have 10 years experience in the exact job they're hiring for so I really don't know what they're looking for or if they're looking that hard.
Readit News
Who knows what we could find. So many books have been lost.
The ancient Roman elite often had extensive personal libraries which they shared with their friends, almost like a very primitive book publishing industry.
The history of the Roman Empire can be divided into two overarching periods - the pagan era and the Christian era. A pagan library is significantly more valuable than a Christian library, as we have far fewer pagan texts than Christian ones, and the few we do have were often corrupted by later Christian copyists. The word pagan is doing meaningful work in the sentence.
I have decided that I, too, shall use obscure things as benchmarks and references. It's pretty good fun. In this post-Ragnarok-Online world one can imagine we need more such milestones to judge other things by.
One of the reasons only a fraction of a percent of the classical texts reached our days is the fact that Christians suppressed those texts, directly (by destroying them) and indirectly (by closing the libraries and temples and institutions of learning which preserved those texts).
Sure you could argue the terminology is very christian-centric, perhaps even offensive to pre-christian romans, but quite frankly that's a very uninteresting debate compared to the topic at hand.
Can you provide some citations on the technology being used in situ without digging up? As far as I understood this is the application of technology widely popularized by the Herculaneum Challenge, where scrolls are still physically dug up, and x-rayed (which will slowly still damage the scrolls) but without physically breaking them open as was repeatedly attempted in the past.
I don't care much about the slow damage from x-rays: as long as the content is succesfully extracted, one can imagine little other use for the scrolls as is.
I mostly hope some lost works on mathematics will be recovered..
So they have the scans of the rolled up scrolls, this is "just" (ha!) using the scan data with lots of algorithms and compute (AI? I presume so) to virtually unroll the scrolls and read the ink off the page.
You are right about not unrolling them though. Many scrolls were destroyed in previous attempts to unroll them physically, so it is fascinating to see how the technology has progressed to allow reading without unrolling.
https://www.youtube.com/watch?v=cw44V49Fz9U
So much for the global internet.
https://www.finebooksmagazine.com/fine-books-news/inside-her...
In particular this part:
Researchers are further refining the image using a new segmentation approach in the hopes that it will improve the coherence and clarity of the lines of text currently visible, and perhaps reach the end of the papyrus, the innermost part of the carbonised scroll, where the colophon with the title of the work may be preserved.
So the new article is indicating they were able now to decipher the title, and also indicates maybe why the title was not the first thing deciphered (presumably it is hardest to read the innermost parts.)
I'm curious why the title is in the inside of the scroll. That implies you have to completely open it to read the title - is that the way scrolls are usually written?
You have 2 teams using the same data, getting to the same conclusion. You also have an author that's known from other sources, with writings that we already "have". Then you have a team of experts reviewing this. Chances are these are real findings and not "hallucinations". Not everything in ML is gen-ai...
All interpretation of ink as Greek letters is done purely by human inference. This may lead to errors, especially in parts where the ink is preserved especially poorly or where the text is totally different from expectations, but it would be classic human error instead of AI hallucination.
> You also have an author that's known from other sources, with writings that we already "have".
> Then you have a team of experts reviewing this.
Only the first of those points is evidence against the result being hallucinated.
Well, that's exactly what you'd expect from a hallucination no? If the model is overfit enough on the relevant corpus, a title that already exists should be much more likely.
https://en.wikipedia.org/wiki/Philodemus
Neat
As a history nerd and jaded software developer, I've been wondering a lot lately how I can use my tech skills for archeological research. Is there any way for someone with most of a bachelors to get into this kind of thing?
[0] https://scrollprize.org/jobs