26 ms·

> houses around 130,000 clay tablets from ancient Mesopotamia > Since so few scholars can read the languages on the tablets—the overwhelming majority remain unt

by nico 2y ago

> houses around 130,000 clay tablets from ancient Mesopotamia > Since so few scholars can read the languages on the tablets—the overwhelming majority remain untranslated This would be an awesome job for AI. Get these scholars to train the AI, then have it automatically translate all the other tablets Also, are there online scans of the tablets? Maybe the problem is not the scarcity of scholars, but gatekeeping? On a tangent: > These tablets, which likely come from Sippar, a Babylonian city in modern Iraq, were acquired by the museum more than a century ago and date to between 1900 and 1600 B.C. Is "acquired" a euphemism for stolen here?

mangamadaiyan 2y ago
How would we know that the AI didn't hallucinate when it translated a tablet? In other words, how could one trust the translation? A human translator can still provide some justification for their interpretation. Also, "acquired" could simply mean that the museum paid money to a third party to, well, acquire the piece(s) in question while not inquiring too deeply into how said third party came into possession of the objects that they sold.
- bruce511 2y ago
  At the very least one could see from the AI if the tablet is interesting or not. That could triage tablets for human translation.
  - mangamadaiyan 2y ago
    Sure, as long as one has some means of predicting the probability of false positives as well as false negatives. Until then, colour me unconvinced of the (f)utility of this approach :)
    - timschmidt 2y ago
      LLMs do a pretty darn good job of translating other languages, even preserving inflection and tone and rhyme in some cases. Same when translating programming languages. If the training pool is large enough, they should be quite good at it. No problem marking them as machine translated and keeping track of which have been spot-checked by experts either.
      - Keysh 2y ago
        I think "if the training pool is large enough" is a real issue here. We're not talking about living languages with large, properly attested and annotated corpuses. Indeed, one of the thing you'd probably like the translators to do is identify rare or unique words that can be added to our existing knowledge of these languages.
        timschmidt 2y ago
        > I think "if the training pool is large enough" is a real issue here. It would be really neat to set up something like a wiki populated with the existing translations and machine translations done via LLM, and to periodically re-train the LLM on all the newly manually verified translations and automatically re-run the machine translations after. The whole thing could move incrementally toward high quality output.
  - Groxx 2y ago
    This is the same thing as trusting its translation though. Or worse, since there's more "interpretation" happening (as it's basically a summary rather than a whole text, and you can't trust either of those steps).
    - flir 2y ago
      It's like using a classifier to sort biopsies from most-likely to least-likely to be Really Bad News. A human will get to them all eventually, but you can at least fast-track the ones that are almost certainly Really Bad News. (I think I just said "triage" in a lot more words, to be honest).
- wood_spirit 2y ago
  I remember a time team episode where they found some Ogham script on a stone in an early Christian church now on a fairway on the isle of Mann. They got photos to the leading professor in Ogham who translated it for them. I happened to mention this on HN a few months ago in relation to something or other, and a commenter replied to me to explain that that translation was no longer sound, and the current understanding was that the tablet said something completely different instead! (Found the link they gave! https://www.babelstone.co.uk/Blog/2008/05/throng-of-fifty-warriors-routed-by.html https://www.babelstone.co.uk/Blog/2008/05/throng-of-fifty-wa...)
  - YeGoblynQueenne 2y ago
    If it wasn't an LLM that found the professor's mistake then the point I think you're trying to make is, well, missing the point: a human made a mistake, and humans found out and corrected it. The question is what happens when LLMs make mistakes. Will people still be careful enough to catch and correct them, as often as we find and correct the mistakes we make ourselves? Or will our ability to do so be overwhelmed by the extreme rate at which LLMs can generate text?
    - wood_spirit 2y ago
      My point was more meta - the experts don’t have a fixed opinion on the translations of rare ancient texts in the first place, ergo there is nothing to train the llma on.
      - YeGoblynQueenne 2y ago
        Thanks for clarifying. But note that the same applies to modern translations. There is no agreement between experts (translators) what it means for a translation to be "good" or "bad", or anything in between. So the done thing in translation is to choose some existing translation as a "gold standard" and use that one, without any assumptions of how good or bad it is. Sometimes there's an attempt to rate translations by polling humans but that can not be easily done at scale and certainly not at web-scale, as in LLMs (the modern de facto standard for automatic translation as far as I can tell; I haven't been watching the field closely lately). The same logic is applied to metrics like BLEU and ROUGE scores, used to measure the goodness of a translation. The general idea is that you choose a translation to be the "gold standard" and then compare the n-grams in the gold standard, and the automatic translation, for overlap. It's a very crude and imperfect measure and it's one reason why, in practice, we have no idea how good automatic translations are especially now that automatic translation systems are deployed and working every day with millions of input and output texts that nobody can reasonably be expected to evaluate. In any case just because there's no fixed opinion on what is a good translation, doesn't mean an LLM can't be used to produce one. It probably will. Assuming I understand your point better now, I agree that this is going to cause trouble down the line.
dghf 2y ago
> Get these scholars to train the AI, then have it automatically translate all the other tablets Which is great until it blithely translates an ancient curse and initiates the apocalypse.
- timschmidt 2y ago
  Klaatu barada... necktie... neckturn... nickle... noodle... It's an N word, definitely an N word.
  - technothrasher 2y ago
    That's not a curse, that's a 'stand down' instruction. I would be much happier to see that, than, say, "Ph'nglui mglw'nafh Cthulhu R'lyeh wgah'nagl fhtagn".
    - dylan604 2y ago
      Unless as protection against hacking attempts at issuing the stand down command initiates a curse. After 3 attempts at the phrase, you must wait 30 seconds before your next attempt. After 5 attempts, you are personally frozen. After 10 attempts, all stand down instructions are wiped and the world is screwed at that point.
- InsideOutSanta 2y ago
  You mean that hasn't already happened? What the hell is going on then?
  - ben_w 2y ago
    Just the usual stuff, unfortunately. Given how chaotic people are, it's remarkable that millions of us working together isn't invariably a constant raging dumpster fire, and instead is only a bit of smoke here and there that sometimes flares up a bit.
    - 2y ago
graemep 2y ago
> Is "acquired" a euphemism for stolen here? Given what and where they were from it almost certainly means donated or sold to the museum by the archeologists who excavated them in the first place.
- Keysh 2y ago
  It sounds like they were excavated in 1881-1882 (e.g., https://en.wikipedia.org/wiki/Sippar#Archaeology https://en.wikipedia.org/wiki/Sippar#Archaeology) by Hormuzd Rassam ("widely believed to be the first-known Middle Eastern and Assyrian archaeologist from the Ottoman empire."; https://en.wikipedia.org/wiki/Hormuzd_Rassam https://en.wikipedia.org/wiki/Hormuzd_Rassam) working in conjunction with the British Museum. "From 1877 to 1882, while undertaking four expeditions on behalf of the British Museum, Rassam made some important discoveries. Numerous finds of significance were transported to the museum, thanks to an agreement made with the Ottoman Sultan by Rassam's old colleague Austen Henry Layard, now Ambassador at Constantinople, allowing Rassam to return and continue their earlier excavations and to 'pack and dispatch to England any antiquities [he] found ... provided, however, there were no duplicates.' A representative of the Sultan was instructed to be present at the dig to examine the objects as they were uncovered." So, not a euphemism for "stolen".
- tbrownaw 2y ago
  Yes but can everyone involved - both the original archeologists who excavated them and the people who bought them - claim ancestry from whatever ancient society made them, or at least the oldest traceable group to have resided on that land? AIUI that is how "stolen" tends to be construed for archaeological things these days.
  - graemep 2y ago
    That idea makes sense in an American context where there is a clear divide between indigenous and more recent groups. The same is true of some other places like Australia. In Eurasia and Africa this does not work. "The oldest traceable group" may be many invasions, conquests and mass migrations after the things were made. The current residents or earliest traceable group may well have enslaved, oppressed, or killed the group who made the things. In many, many cases they would have erased the culture that made them. In the case of these tablets there is no one who can meaningfully claim to be an Akkadian.
mda 2y ago
See aicuneiform.com
idoubtit 2y ago
> Also, are there online scans of the tablets? Maybe the problem is not the scarcity of scholars, but gatekeeping? With a few minutes of searching the internet, you would not have written that inappropriate question. Tablets are 3D objects which are most often eroded and broken. A plain photography, like one would make for a sheet of paper, is useless in the general case. To bring out the cuneiform characters, the light must be raking the surface. And don't forget that that surface can be concave, and text can go over the sides. Most tablets need many photos with different lights. It's a long and hard work that is not automated. Guess what, Scholars have heard about AI. If AI could help them publish astonishing papers and push forward their career, do you really think they wouldn't rush to it?
cess11 2y ago
Have you ever read translations of such ancient texts? Why are you so certain there are enough reliable translations to do the model training? Look at https://etcsl.orinst.ox.ac.uk/cgi-bin/etcsl.cgi?text=t.3.1.21 https://etcsl.orinst.ox.ac.uk/cgi-bin/etcsl.cgi?text=t.3.1.2... , for example. To me it doesn't come across as something I would expect an LLM to be able to produce. They are basically decent at translating between certain languages, and only documents that are very formal. As soon as metaphor, idioms and so on come up in source texts they suck. There's more translated sumerian you can read here: https://etcsl.orinst.ox.ac.uk/edition2/etcslbycat.php https://etcsl.orinst.ox.ac.uk/edition2/etcslbycat.php
- starluz 2y ago
  [dead]

by nico 2y ago

mangamadaiyan 2y ago

How would we know that the AI didn't hallucinate when it translated a tablet? In other words, how could one trust the translation? A human translator can still provide some justification for their interpretation. Also, "acquired" could simply mean that the museum paid money to a third party to, well, acquire the piece(s) in question while not inquiring too deeply into how said third party came into possession of the objects that they sold.

bruce511 2y ago
At the very least one could see from the AI if the tablet is interesting or not. That could triage tablets for human translation.
- mangamadaiyan 2y ago
  Sure, as long as one has some means of predicting the probability of false positives as well as false negatives. Until then, colour me unconvinced of the (f)utility of this approach :)
  - timschmidt 2y ago
    LLMs do a pretty darn good job of translating other languages, even preserving inflection and tone and rhyme in some cases. Same when translating programming languages. If the training pool is large enough, they should be quite good at it. No problem marking them as machine translated and keeping track of which have been spot-checked by experts either.
    - Keysh 2y ago
      I think "if the training pool is large enough" is a real issue here. We're not talking about living languages with large, properly attested and annotated corpuses. Indeed, one of the thing you'd probably like the translators to do is identify rare or unique words that can be added to our existing knowledge of these languages.
      - timschmidt 2y ago
        > I think "if the training pool is large enough" is a real issue here. It would be really neat to set up something like a wiki populated with the existing translations and machine translations done via LLM, and to periodically re-train the LLM on all the newly manually verified translations and automatically re-run the machine translations after. The whole thing could move incrementally toward high quality output.
- Groxx 2y ago
  This is the same thing as trusting its translation though. Or worse, since there's more "interpretation" happening (as it's basically a summary rather than a whole text, and you can't trust either of those steps).
  - flir 2y ago
    It's like using a classifier to sort biopsies from most-likely to least-likely to be Really Bad News. A human will get to them all eventually, but you can at least fast-track the ones that are almost certainly Really Bad News. (I think I just said "triage" in a lot more words, to be honest).
wood_spirit 2y ago
I remember a time team episode where they found some Ogham script on a stone in an early Christian church now on a fairway on the isle of Mann. They got photos to the leading professor in Ogham who translated it for them. I happened to mention this on HN a few months ago in relation to something or other, and a commenter replied to me to explain that that translation was no longer sound, and the current understanding was that the tablet said something completely different instead! (Found the link they gave! https://www.babelstone.co.uk/Blog/2008/05/throng-of-fifty-warriors-routed-by.html https://www.babelstone.co.uk/Blog/2008/05/throng-of-fifty-wa...)
- YeGoblynQueenne 2y ago
  If it wasn't an LLM that found the professor's mistake then the point I think you're trying to make is, well, missing the point: a human made a mistake, and humans found out and corrected it. The question is what happens when LLMs make mistakes. Will people still be careful enough to catch and correct them, as often as we find and correct the mistakes we make ourselves? Or will our ability to do so be overwhelmed by the extreme rate at which LLMs can generate text?
  - wood_spirit 2y ago
    My point was more meta - the experts don’t have a fixed opinion on the translations of rare ancient texts in the first place, ergo there is nothing to train the llma on.
    - YeGoblynQueenne 2y ago
      Thanks for clarifying. But note that the same applies to modern translations. There is no agreement between experts (translators) what it means for a translation to be "good" or "bad", or anything in between. So the done thing in translation is to choose some existing translation as a "gold standard" and use that one, without any assumptions of how good or bad it is. Sometimes there's an attempt to rate translations by polling humans but that can not be easily done at scale and certainly not at web-scale, as in LLMs (the modern de facto standard for automatic translation as far as I can tell; I haven't been watching the field closely lately). The same logic is applied to metrics like BLEU and ROUGE scores, used to measure the goodness of a translation. The general idea is that you choose a translation to be the "gold standard" and then compare the n-grams in the gold standard, and the automatic translation, for overlap. It's a very crude and imperfect measure and it's one reason why, in practice, we have no idea how good automatic translations are especially now that automatic translation systems are deployed and working every day with millions of input and output texts that nobody can reasonably be expected to evaluate. In any case just because there's no fixed opinion on what is a good translation, doesn't mean an LLM can't be used to produce one. It probably will. Assuming I understand your point better now, I agree that this is going to cause trouble down the line.

dghf 2y ago

> Get these scholars to train the AI, then have it automatically translate all the other tablets Which is great until it blithely translates an ancient curse and initiates the apocalypse.

timschmidt 2y ago
Klaatu barada... necktie... neckturn... nickle... noodle... It's an N word, definitely an N word.
- technothrasher 2y ago
  That's not a curse, that's a 'stand down' instruction. I would be much happier to see that, than, say, "Ph'nglui mglw'nafh Cthulhu R'lyeh wgah'nagl fhtagn".
  - dylan604 2y ago
    Unless as protection against hacking attempts at issuing the stand down command initiates a curse. After 3 attempts at the phrase, you must wait 30 seconds before your next attempt. After 5 attempts, you are personally frozen. After 10 attempts, all stand down instructions are wiped and the world is screwed at that point.
InsideOutSanta 2y ago
You mean that hasn't already happened? What the hell is going on then?
- ben_w 2y ago
  Just the usual stuff, unfortunately. Given how chaotic people are, it's remarkable that millions of us working together isn't invariably a constant raging dumpster fire, and instead is only a bit of smoke here and there that sometimes flares up a bit.
  - 2y ago

graemep 2y ago

> Is "acquired" a euphemism for stolen here? Given what and where they were from it almost certainly means donated or sold to the museum by the archeologists who excavated them in the first place.

Keysh 2y ago
It sounds like they were excavated in 1881-1882 (e.g., https://en.wikipedia.org/wiki/Sippar#Archaeology https://en.wikipedia.org/wiki/Sippar#Archaeology) by Hormuzd Rassam ("widely believed to be the first-known Middle Eastern and Assyrian archaeologist from the Ottoman empire."; https://en.wikipedia.org/wiki/Hormuzd_Rassam https://en.wikipedia.org/wiki/Hormuzd_Rassam) working in conjunction with the British Museum. "From 1877 to 1882, while undertaking four expeditions on behalf of the British Museum, Rassam made some important discoveries. Numerous finds of significance were transported to the museum, thanks to an agreement made with the Ottoman Sultan by Rassam's old colleague Austen Henry Layard, now Ambassador at Constantinople, allowing Rassam to return and continue their earlier excavations and to 'pack and dispatch to England any antiquities [he] found ... provided, however, there were no duplicates.' A representative of the Sultan was instructed to be present at the dig to examine the objects as they were uncovered." So, not a euphemism for "stolen".
tbrownaw 2y ago
Yes but can everyone involved - both the original archeologists who excavated them and the people who bought them - claim ancestry from whatever ancient society made them, or at least the oldest traceable group to have resided on that land? AIUI that is how "stolen" tends to be construed for archaeological things these days.
- graemep 2y ago
  That idea makes sense in an American context where there is a clear divide between indigenous and more recent groups. The same is true of some other places like Australia. In Eurasia and Africa this does not work. "The oldest traceable group" may be many invasions, conquests and mass migrations after the things were made. The current residents or earliest traceable group may well have enslaved, oppressed, or killed the group who made the things. In many, many cases they would have erased the culture that made them. In the case of these tablets there is no one who can meaningfully claim to be an Akkadian.

mda 2y ago

See aicuneiform.com

idoubtit 2y ago

> Also, are there online scans of the tablets? Maybe the problem is not the scarcity of scholars, but gatekeeping? With a few minutes of searching the internet, you would not have written that inappropriate question. Tablets are 3D objects which are most often eroded and broken. A plain photography, like one would make for a sheet of paper, is useless in the general case. To bring out the cuneiform characters, the light must be raking the surface. And don't forget that that surface can be concave, and text can go over the sides. Most tablets need many photos with different lights. It's a long and hard work that is not automated. Guess what, Scholars have heard about AI. If AI could help them publish astonishing papers and push forward their career, do you really think they wouldn't rush to it?

cess11 2y ago

Have you ever read translations of such ancient texts? Why are you so certain there are enough reliable translations to do the model training? Look at https://etcsl.orinst.ox.ac.uk/cgi-bin/etcsl.cgi?text=t.3.1.21 https://etcsl.orinst.ox.ac.uk/cgi-bin/etcsl.cgi?text=t.3.1.2... , for example. To me it doesn't come across as something I would expect an LLM to be able to produce. They are basically decent at translating between certain languages, and only documents that are very formal. As soon as metaphor, idioms and so on come up in source texts they suck. There's more translated sumerian you can read here: https://etcsl.orinst.ox.ac.uk/edition2/etcslbycat.php https://etcsl.orinst.ox.ac.uk/edition2/etcslbycat.php

starluz 2y ago
[dead]