How AI could help us with ancient languages like Sumerian | Hacker News Reader

How AI could help us with ancient languages like Sumerian

bbc.com

123 pointsnkurz44 comments

I tracked down a book earlier this year (via inter library loan) called Archaic Bookkeeping about some of the oldest engravings found, mostly pre-cuneiform. This was essentially the invention of arithmetic, accounting, and writing all at once. It’s amazing stuff. The book was written in the early 90s and touched on the earliest attempts to have computer assistance in translation, done throughout the 80s. It would be great to see what today’s techniques could do.

People use the 'set in stone' phrase all the time, it puts a different spin on it when 6000 years later, we are still analyzing some poor guys tax returns. This is fascinating stuff, too bad data mining the mundane lives of social media users pays better.

Can't sell a Sumerian a new model iPhone.

And to play devil's advocate here, what does this old data potentially get us? Like OP said, it may just be some dude's tax info. Nice to have, and maybe useful for painting a picture of day-to-day life, but that's not going to do anything for climate change, stagnant wages, or Mutually Assured Destruction?

If I'm not mistaken, generalising your question leads to: "What is the use of studying history?"

Attitudes like the one expressed in the comment you're replying to is exactly why STEM majors should have a required humanities/liberal arts component.

I've seen this attitude in engineering majors who have been required to take humanities and liberal arts classes. Maybe such a requirement helps some people, but I suspect resentment is a more common outcome amongst people already predisposed to devalue the humanities.

I've seen that happen personally, but I think that's a failure of the instructor or the system (or both), not of the subject matter itself. The liberal arts should be fascinating if taught in a compelling way. It's the story about what it means to be a person, which is the only universal activity of humans both present and past.

Unfortunately, many don't seem to see the point of things as learning "what it means to be a person". They want everything to boil down to a nice engineering problem that, while complex, is solvable with one solution being better/easier. I remember reading an HN post about "Not everything is an engineering problem" and, if I'm not mistaken, the top comment was basically "Like hell it isn't." That's the bigger issue, imo, the mindset we've pushed into people about STEM being the end-all-be-all, and how it's the One True Way, or whatnot. We have, as a group, completely devalued the liberal arts and now are seeing the effects of that and wondering why people don't see the need for them (and I say that as a former STEM-Lord convert).

As a physics major who the requirements helped, especially my philosophy courses (an intro one and a Philosophy of Science course), I think a lot of it comes down to teachers, and how the material is presented. It was never presented to me as boring, and the teachers never tried to claim it was more important than science or the end-all-be-all of the future. And all of that eventually led me down the road where I'd rather do more work in the humanities than anything science related, and would love to do humanities/liberal arts research professionally.

That said, I think some of the problem is the mindset STEM people are kinda forced into -- one solution, and it has to be a scientific one. It makes the whole world boil down into an engineering problem, and anyone who doesn't like it/doesn't agree can be damned because "look at what science has done"/

I'd be more enthusiastic about such requirements if humanities/lib-arts courses actually provided a solid answer to how one should live and what one should want. But they don't. They provide lots of different answers, with no great way to choose between them. Going through a mass of required reading only to reach the conclusion that living for bling+bitches is actually a pretty OK idea doesn't seem like a good use of time.

>They provide lots of different answers, with no great way to choose between them.

...that's the point.

The entire purpose of a liberal arts education is to learn about what is important to you personally, what it means to be a person in a world full of other people making that same journey, how to understand the events that lead to where we are today, how to think about and approach new ideas, how to evaluate and process new information and so on. Your comment that shows a misunderstanding of the liberal arts reflects a core failure of the education system that is unfortunately all too common.

"...only to reach the conclusion that living for bling+bitches is actually a pretty OK idea."

Actually, essentially all of the suggestions agree that living for "bling+bitches" is a pretty poor idea. For different reasons, of course.

Hedonism is respectable enough that textbooks in ethics keep presenting it. The other ethical approaches kick it in the shins pretty hard, but no one seems to have a decisive refutation. Ethicists from other traditions just think hedonists are punks.

This is true. On the other hand, many of the approaches that get kicked around as "hedonism", like Epicurus, are pretty far from "bitches+bling".

Rather, why there should be more humanities / liberal arts outreach (and funding), so that STEM majors are more likely to have access to, engage with, and fall in love with them.

I understand the impulse behind such comments, but making something "required" is generally a poor way to get people to love or appreciate something. (Think of all the people who "hate mathematics".) In education, often the system is broken, good instructors are to find, etc. (Consider Lockhart's Lament, and its generalization: https://www.scottaaronson.com/blog/?p=410.) We see this in politics too (I don't mean just electoral politics, but e.g. open-source communities with codes of conduct etc), where even simple basic human kindness, when made mandatory, often seems to provoke backlash. We can't require people to be nice, and we can't require people to appreciate the humanities. We have to do the hard work of gently persuading.

Anything made mandatory always becomes unpleasant.

Consider sex, sounds fun when spontaneous and free, but slave-staffed prostitution is not so fun from the slave's perspective.

Consider food, how could a nice meal be unpleasant, but force feeding has been used for torture, not so cool anymore.

Consider religion, or politics, or housework, etc...

To some extent if you're allowing people into higher ed who are not higher-quality people, they're not going to appreciate a wide ranging education and a wide education is useless for most vocational training (engineering, programming, etc) anyway. Historically higher vocational training was only pursued by the highest class of people; not so much anymore; probably time to remove legacy nobility from higher ed for the masses.

Are you genuinely comparing educational requirements to slavery and torture?

Generalizing that question leads to "How should we live, knowing that we will die?"

Knowing one's past is knowing one's self.

"The past is never dead. It's not even past.*"

William Faulkner. Or possibly Yogi Berra.

To put it more geekly: it's like uncovering an old git commit from a dead branch. Is it useful? It's only useful to try understand where things came from. Is that useful? In the grand scheme of things if we don't know where we're coming from we don't have any reliable way to asses where we're heading towards. In addition, the exercise sharpens our existing tools.

I had an illuminating conversation with a friend once. We were contrasting the values of our respective areas of interest (her - business, me - research) and I realised something odd.

To her, non-commercial research is a nice luxury we can afford in addition to the principal purpose of society, advancing material condition.

To me, material advance is mostly the necessary price we must pay to accomplish what's truly important - advancing our understanding of everything around us.

Obviously I'm simplifying somewhat, research is often necessary for material advance and I'm not about to trade in my creature comforts, but the two worldviews are fundamentally at odds, and trying to understand the purpose of one in terms of advancing the other is fruitless.

Research into history isn't designed to answer your priorities. But to those interested in it, it's the fundamental point, for which addressing your priorities is merely necessary bookkeeping.

Thanks for taking the time, that is an illuminating point.

I remember spending countless hours arguing with someone who genuinely believed 'greed is good', a total waste of time, no progress whatsoever on either side.

Thinking about it now, my sparring partners attitude makes some sense, if 'advancing our material condition' is indeed the end game. It's like two engineers arguing about implementation details, without realizing they are designing different apps.

Buzzwords like 'Goal Oriented' or 'Understand your clients motivation' fly around, but your point gives them more dimension. Thanks.

There is also a great deal of lost history recorded on (some of) those tablets. I do find that quite valuable. So much history has been lost, and destroyed. ISIL/ISIS actually destroyed a lot in the last few years in this area. I find this INCREDIBLY sad. https://en.wikipedia.org/wiki/Destruction_of_cultural_herita...

"A tablet with beer allocations – a good 5,000 years old – pinned up in a glass case like a butterfly for all to see. What can be interesting in that beyond the fact, obvious enough when you think about it, that hard-working individuals throughout history have always wanted their glass of beer?"

http://www.bbc.co.uk/ahistoryoftheworld/objects/TnAQ0B8bQkSJ...

Irving Finkel, mentioned in the article and quite a character (check out Youtube videos with him; the latest BM Curator's Corner has him teaching someone cuneiform and then suggesting they leave their clay tablets on the floor of the Ashurbanipal exhibit to see who panics).

NLP(NLU, NLI, NLG) is the tip of the spear when it comes to AI and has its roots in the study of Epigraphy [1] which is essentially the study of ancient inscriptions and symbols. Analyzing human language and thought surrounding objects is key to advances in AI.

The study of epigraphy is the microscope while the study of human aging, extending human lifespan is the telescope that lets us see the stars in the future. With Epigraphy you can change dates in human history which changes a lot of what we thought we knew about ourselves.

The best NLP is based on principles that underpin epigraphy combined with neuroscience and research in cognition. The interesting part is that unbeknownst to most, the study of the past (epigraphy/NLP) is being used to solve for the future (space biosciences - LET radition, DNA repair, extending human lifespan making real space travel possible for the purpose of inhabiting other planets). Sounds out there, but this is what we work on at LBNL/DOE/DARPA/NASA related to predicting new interactions between proteins, genes etc.

Fun fact: Epigraphy, ciphers/deciphering, info security, cryptography are all related [2]

[1] https://en.wikipedia.org/wiki/Epigraphy

[2] From the epigraphy mailing list:

"Georges Depeyrot <Georges.depeyrot@ehess.fr>

To: epigraphy@yahoogroups.com

Sep 20, 2003 at 10:28 PM

Dear List members, We are publishing books on numismatics and economy (see Moneta <http://www.cultura-net.com/moneta/content.htm>). I just want to inform you that one of the books in press is devoted to "coins, money and epigraphy, 1st - 3rd c." (author, Pr St. Mrozek, Uty of Gdansk, Poland). It will concern the Roman Empire."

cuneiform has a lot of data, i would be more interested though in things that people claim to have translated already, like for instance Egyptian hieroglyphs where theres a solid body convinced it's all correctly translated ,but some parties dispute some basic symbols which change the meanings of texts dramatically.

Skynet trained on ancient Sumerian texts... I like the potential storyline.

Neal Stephenson is proven right one more time...

This is partially what it was implied David from Prometheus was doing at the start of the film (although he appeared to be working back from modern pronunciation to earlier proto-languages too)

Yes! I'm pretty sure it was based off of Sitchin's translations of the tablets, and David was looking for the alien race that created the human hybrids (us). It's actually very close to the Adam and Eve story, like how the flood tale in the tablets mirror the Great Flood in the Bible. https://en.wikipedia.org/wiki/Zecharia_Sitchin

How AI could help us with anything, really.

AI approaches to "reading" the Voynich manuscript have not been terribly successful.

In this case, the script is not a problem. The article is about bulk machine translation of a lot ancient texts.

And unlike the Voynich manuscript, here we have known good translations to start from.

It's probably because it's most likely nonsense and can't be read.

It likely is nonsense, given that its symbols' frequencies don't match up with other natural (or constructed, for that matter) language word distributions. Either the manuscript is encrypted and salted, or it's a big fat phony.

Sumerian has already been translated, so it’s a labor saving problem.

Perhaps that is yet another indication that the Voynich Manuscript is a hoax.

In addition to the Cuneiform Digital Library Initiative, there is also the Electronic Text Corpus of Sumerian Literature, http://etcsl.orinst.ox.ac.uk/. (A nice example: "The advice of a supervisor to a younger scribe", transliteration: http://etcsl.orinst.ox.ac.uk/cgi-bin/etcsl.cgi?text=c.5.1.3&..., translation: http://etcsl.orinst.ox.ac.uk/cgi-bin/etcsl.cgi?text=t.5.1.3 .)

There are three problems with machine translation of cuneiform texts (Problem 0 is accumulating enough pieces of a text to make it worthwhile[1]):

The first is reading the signs on the tablet, which is probably incomplete and damaged. Further, some scribes weren't all that spectacular; IIRC, the early stages of Assyriology benefited because of the early discovery of Ashurbanipal's library at Ninevah, where he'd put some money into ensuring the tablets were good quality.

I believe there has been some work on 3D scanning of tablets, but I haven't found any major collection of them. Or of high quality photos, for that matter.

The second is transliteration of the signs into usable text. The article mentioned image recognition for cylinder seals, but I'd be interested in any significant success at recognizing text.

The third is translation, described in the article. This is probably the easiest; Akkadian (used in Assyria and Babylon) and Sumerian are both understood reasonably well.

I think there's two useful results of this research:

The general "digital humanities" giant database of people/places/things: transactions, prices, names, etc. It's possible with that to understand Mesopotamian civilization better than we do Greek or Roman.

And identifying, of the giant mass of archived tablets, those that really do deserve a closer look. Like Irving Finkel's translation ("discovery" isn't the right word) of another copy of Atrahasis' flood story.

[1] There's a cute story about Cyrus the Great's cylinder, I believe; one copy of it was broken either in antiquity or around the time of discovery and transportation. The two pieces wound up in Europe and America (Penn State?) until somebody very observant realized they were two parts of the same cylinder. And I'm given to believe that the British Museum has employed people to re-jigsaw the gravel at the bottom of the incoming crates of tablets back into something useful.

Maybe Sumerian and Enochian are best left unspoken.

Enochian isn't a real language, though, it's just something John Dee made up while tripping on mercury or something.

Sumerian is a pretty cool language. https://www.youtube.com/watch?v=8H-Vhsma7ck

> Hey, Hiro, the black-and-white guy says, you want to try some Snow Crash?

Reply on news.ycombinator.com