|
Judgy Fucker posted:It's still valuable as literature, though. And if nothing else it's really drat exciting to get access to previously-lost ancient texts. Wake me up when they find Lives.
|
# ? Apr 30, 2024 16:07 |
|
|
# ? May 2, 2024 22:32 |
|
Silver2195 posted:To be clear, this would have been written long enough after Plato's death that it's historical accuracy is questionable. All texts have questionable historical accuracy! People still argue about whether Plato’s eyewitness portrait of Socrates’ teaching is accurate or not. Aubrey’s “Brief Lives” posted:What uncertainty doe we find in printed histories? they either treading too neer on the heeles of trueth that they dare not speake plaine, or els for want of intelligence (things being antiquated) become too obscure and darke!
|
# ? Apr 30, 2024 16:30 |
|
Man if we find an Etruscan to Latin translation that would be huge.
|
# ? Apr 30, 2024 16:37 |
|
WoodrowSkillson posted:Man if we find an Etruscan to Latin translation that would be huge. All the toasty scrolls are Greek apparently, but now that there is a chance of reading them, maybe Latin stuff will be dug up?
|
# ? Apr 30, 2024 16:41 |
|
skasion posted:All the toasty scrolls are Greek apparently, but now that there is a chance of reading them, maybe Latin stuff will be dug up? I think the style at the time was to have separate Greek and Latin libraries. At least we can say that one hasn't been dug up yet and there's a lot of places left to be excavated, so here's hoping! Man, the whole "reading burnt ancient scrolls with really thin laser beams" thing makes me remember being excited about the possibilities of future technology instead of just dreading what they come up with next.
|
# ? Apr 30, 2024 18:45 |
|
barbecue at the folks posted:I think the style at the time was to have separate Greek and Latin libraries. At least we can say that one hasn't been dug up yet and there's a lot of places left to be excavated, so here's hoping! I'm afraid to tell you that really thin laser beams are an important component in the Torment Nexus
|
# ? Apr 30, 2024 23:59 |
|
barbecue at the folks posted:I think the style at the time was to have separate Greek and Latin libraries. At least we can say that one hasn't been dug up yet and there's a lot of places left to be excavated, so here's hoping! We have a lot of missing texts, right? What if we feed the existing ones into an AI machine learning neurally networked algorithm AI?
|
# ? May 1, 2024 00:12 |
|
I don't doubt that once the training data exists, machine learning will be an important part of decoding the texts more rapidly (and possibly already is), but existing models basically completely failed to make any headway without application-specific training. The initial breakthrough came from a guy just staring at the images until the vague squiggles started looking like greek
|
# ? May 1, 2024 01:08 |
This isnt the same project I think, but yea plenty of people are trying to use machine learning vision stuff to read the scrolls. See the Scroll Prize website. Generally, the problem is that carbonized scrolls are a 3D medium of difference sheets of varying thickness rolled up together. So you've got to be able to detect both the 3D blobs of ink on a piece of papyrus, but also be able to seperate out the different sheets from your 3D scroll into 3D sheets. And you're dealing with huge data volumes even for a single scroll, because of how high resolution the scanning data needs to be to detect the ink in the carbonized scroll
|
|
# ? May 1, 2024 01:16 |
|
Has there been any progress on Linear A / the Harappan script? It feels wild to me there's all this text we still can't actually read even with all the technological headway these days.
|
# ? May 1, 2024 04:19 |
|
Can’t read the Indus script either yet. That’s a big one .
|
# ? May 1, 2024 04:22 |
|
my understanding is that every case of deciphering a lost script happened with a language that was still being spoken at the time of decipherment. If the language behind linear A is truly gone, then it's probably a lost cause.
|
# ? May 1, 2024 04:31 |
|
I read an article about the process used to read the scrolls, and machine learning is indeed involved in reconstructing the images of the scrolls once another part of the process scans and "unrolls" them. Obviously nothing like the "ai" bullshit that's everywhere now, this isn't going to hallucinate anything (the text wouldn't make any sense if it did).
|
# ? May 1, 2024 04:31 |
|
Please stop saying mahine learning and other buzzwords .
|
# ? May 1, 2024 04:33 |
|
Short of finding a multilingual inscription with Linear A in it it's probably going to stay unreadable. I hope that ends up being wrong but as far as I understand it the problem is there's just nowhere to start. There's been lots of guessing with known languages and none of them have ever turned up any results that make sense.
|
# ? May 1, 2024 04:34 |
|
euphronius posted:Please stop saying mahine learning and other buzzwords . they made up the word machine learning to do the same poo poo they were doing in statistical AI but actually get grants, after the third AI winter or so, as a second round to what they did after the first AI winter. quite the old buzzword you can call it statistics, but the peeps who actually do this sort of thing hate the statisticians at this point and especially their computational incompetence, so it's gotta have another word
|
# ? May 1, 2024 04:40 |
|
I understand RAs have to use those words to get grants / published
|
# ? May 1, 2024 04:41 |
|
The fundamental problem with Linear A and The Indus Valley script is the lack of a large corpus of texts and the lack of any bilingual texts. You need one or the other, preferably both, in order to decipher an unknown script that is writing an unknown language. When it comes to the Indus Valley script, we have a fair amount of texts, 5,509 total known. But they are almost all extremely short. On average, each text has less than four symbols, and 707 unique symbols have been identified. This means it is very hard to find patterns, since the huge number of unique symbols and the extreme brevity of nearly all texts means its quite hard to draw any conclusions. There has been a fair amount of recent computational work on the Indus Valley script, which has been able to show fairly conclusively that is it in fact recording language. This has not always been assumed. Given the brevity of inscriptions, and the huge number of distinct signs relative to the size of the total corpus, people have sometimes suggested it was actually recording something language-adjacent, like personal heraldry or administrative identifications. But thanks to ~statistics~ we are now pretty confident that it is in fact a language. Attempts to use those same computational methods to uncover the underlying language, or find insights into the meaning of the words, have unfortunately been failures. The current evidence simply does not provide the patterns needed to start making conclusions about the underlying language, whether a human is looking at it or a computer is. If we don't find longer texts, or bilingual texts, we will probably never know much more about the Indus Valley script. We are on somewhat more solid ground when it comes to Linear A, since the sign inventory is smaller (around 300), and there's a core group of less than a hundred signs that are more common than others. Crucially, there are also some signs that are shared between Linear A and Linear B, and so we know what those ones mean. This has enabled some basic syntactic conclusions to be drawn, since sometimes we can pick out a few words known from Linear B. The underlying language of Linear A probably had Verb-Subject-Object word order, since in a study of Linear A offering bowls, the word for "olive" (the object) can be identified based on a sign shared with Linear B, the personal name (the subject) can be identified due to its uniqueness, and the verb for "to give" (or something similar) can identified based on it being repeatedly constantly in these offering bowls. That's already more than we know about the Indus Valley script. But the overall corpus of Linear A is fairly small, at around 1,400 total texts, and most of those are texts like offering bowls, short and formulaic. Longer, more varied texts, or bilingual texts, are needed to make more progress on Linear A.
|
# ? May 1, 2024 04:59 |
|
cheetah7071 posted:my understanding is that every case of deciphering a lost script happened with a language that was still being spoken at the time of decipherment. If the language behind linear A is truly gone, then it's probably a lost cause. That's not quite true. One example where a script was deciphered where the underlying language had been totally lost is Meroitic, which was written in Nubia in the late 1st millennium BCE and early 1st millennium CE. The script itself wasn't that hard to work out since its got a number of similarities with Demotic Egyptian, although it also some differences that had to be figured out based on internal evidence, including a totally different system of writing vowels, which took a while to to be fully understood. The Meroitic language is extinct, and it was totally unknown when the first Meroitic inscriptions were discovered in the late 19th century. Meroitic has no known relatives among any other known languages, ancient or modern. (There are people who claim it is related to certain modern Nubian or Nilo-Saharan languages but this is based on very thin evidence. It's probably related to something that it still spoken today but its unclear what). Our knowledge of the Meroitic language is fully reconstructed based on careful study of the texts. Now, the Meroitic language isn't understood that well, and there are a lot of gaps in our knowledge, but everything we know about it we have learned by deciphering texts written in Meroitic. There are a few Egyptian-Meroitic bilingual texts that have helped, but they don't offer that much, since they tend to be short and formulaic. There's no Rosetta Stone equivalent.
|
# ? May 1, 2024 05:09 |
|
euphronius posted:Please stop saying mahine learning and other buzzwords . Machine learning is an actual field, and its application to reading the carbonized Herculaneum scrolls is far more akin to an extremely fancy version of optical character recognition, a mature and sober application of the technology, than to anything about so-called "AI" that gives investors boners lately. It's also completely irrelevant to the question of deciphering scripts that are already perfectly legible.
|
# ? May 1, 2024 05:25 |
|
Bongo Bill posted:Machine learning is an actual field, and its application to reading the carbonized Herculaneum scrolls is far more akin to an extremely fancy version of optical character recognition, a mature and sober application of the technology, than to anything about so-called "AI" that gives investors boners lately. Yeah this is true. Also "machine learning" was deliberately chosen to be a "softer" term than "AI", since the things people were originally calling AI, while very useful, were very far away from what most people would call "intelligence". Tree Bucket posted:We have a lot of missing texts, right? What if we feed the existing ones into an AI machine learning neurally networked algorithm AI? And this is lazy trolling.
|
# ? May 1, 2024 06:33 |
|
I'm not convinced it wasn't supposed to be a joke
|
# ? May 1, 2024 07:27 |
it is a very obvious joke
|
|
# ? May 1, 2024 11:33 |
|
|
# ? May 2, 2024 22:32 |
|
See you'd think so, but I've read the Tech Nightmares thread
|
# ? May 1, 2024 20:57 |