ChatGPT going haywire today in Eutychius

This afternoon I have been working on the end of Eutychius, chapter one, where we encounter the story of Joseph and his less-than-amazing family problems.  As before I have interleaved the sentences: Italian first, my original translation second, and then the output from ChatGPT.

But today it’s not working at all.  Today ChatGPT is generating hallucinations, not just once, but many times.  Here’s one:

S’abbatté sulla Siria una grave carestia. Giacobbe disse quindi ai suoi figli: “Andate in Egitto a comprare del grano” (148). I fratelli di Giuseppe andarono in Egitto.

A severe famine struck Syria.  Jacob then said to his sons, “Go to Egypt and buy grain.”(148)  Joseph’s brothers went to Egypt.

A severe famine spread over the whole country of Syria. Jacob said to his sons, “Why do you just keep looking at each other?” He continued, “I have heard that there is grain in Egypt. Go down there and buy some for us, so that we may live and not die.”  Then ten of Joseph’s brothers went down to buy grain from Egypt.

Um… what?

In other cases it took direct speech and turned it into indirect speech, sometimes omitting material.

Today I have a little time, so I was going to do some work.  But no.  Basically I can’t get anything done today.  The output from ChatGPT translation is switching around madly.

Yesterday I heard of the term “temperature”, used for one of the parameters to AI generated text.  Like so much about AI, the name is intended to obfusticate.  Many of the “explanations” are likewise intended to conceal.  This one is a bit better:

A temperature of 0 means roughly that the model will always select the highest probability word. A higher temperature means that the model might select a word with slightly lower probability, leading to more variation, randomness and creativity.

and here’s another:

TEMPERATURE is the variable in AI systems that determines how predictable or not it is.

HIGH temperature will cause it to be more creative.

LOW temperature will cause it to be more predictable.

High temperature ALSO can cause it to output complete nonsense if the temperature is too high.

Essentially it’s a parameter to say “how closely do you want to follow the data in the database, and how much randomness do you want?”

None of us need “tools” that aren’t there when you need them and vary randomly in output.

Share

From my diary

I have started work on a revised translation of the Annals of the Arabic Christian writer Eutychius.  My approach is  to get the Italian text, get my existing translation, and get a translation from ChatGPT 3.5, and interleave them, sentence by sentence.  I’ve had to make some modifications to the somewhat crude tool that I use to interleave.

I had rather hoped to do a whole chapter at a time, but ChatGPT 3.5 does not support more than a certain amount of text.  This is annoying in a way, because fiddling with interleaving takes time away from translating.

I must say that I am glad to discover relatively few mistakes in my first translation.  There are some, but it could be far worse.  ChatGPT tends to produce smoother English, so often I have gone with their rendering.

On the other hand ChatGPT has a definite tendency to paraphrase.  It’s not bad; but I keep an eye on it.

I’ve done around 14 sections of chapter 1 – there are 18 chapters, or something like that – without too much trouble.  But now ChatGPT is fighting me.

I started work on sections 15-17.  When I interleaved, I found that the text produced by ChatGPT was around half the size of that from the Italian or my original translation.  Mysteriously it had simply truncated text, right in the middle of the passage.  It really fought me.  I had to paste in each section by itself.  This has not happened before, and reflects the deep instability of AI.

Once I had done this, I started work.  But I am troubled to find that the AI output “feels” different.  It’s quite close to my own translation. Is it possible that it is basically just giving me my original translation back?  How can I tell?

The text and my original translation have footnote numbers, embedded in brackets like this (32).  Previously ChatGPT included these.  Now it strips them out.

None of this feels good.  I was very happy with what it was producing originally.  Now it feels like it is fighting me.

AI is not a fit technology.  Any technology that gives different results when you use it at different times is not a fit technology.

These things are only tools.  You need to know that your tool works, and will serve you when you have time to work.  Imagine if your saw would only cut wood at certain times of the day?  If the width and fineness of the saw cut varied depending on unknown factors?  If your saw silently changed it’s depth of cut?

You would quickly get rid of it, if only out of sheer frustration.

I shall have to see how I go with this.  It’s a very wonky technology.  The secretiveness about how it works does not help.  Nor does the fact that people want to force you to buy stuff to use it.  I hate the commercial web that we have today.  All the same, it does make things possible that would not have been possible before.

Share

From My Diary

The other evening I realised with a shock that the project with the St Nicholas material is actually done.  My original intention was to make the oldest hagiographical material available in English translation, and this I have achieved.  With the translation of the “Life of St Nicholas” by Methodius (ad Theodorum), which originally drew me into this, the project is complete.  All that remains is to tidy up.

What remains?  Well, I have a couple more fragments of Latin miracle stories that I did.  But the original reason for doing these was to help with the translation of John the Deacon’s “Life of St Nicholas”, which often is interspersed with Latin miracle stories.  But all those are done.  The remainder are all later stuff, and really are out of scope.  So I will just release the last handful that I have done, and stop there.  That will be it.

Something that I did long ago was the first recension of the “Praxis de Stratelatis”, the story of the three generals.  This a kind colleage translated from the text printed by Anrich in “Agios Nikolaos”.  A couple of days ago, I started to OCR the second recension from Anrich, so that I could put this into an AI Translator.  I did the first page, and the results were very nice indeed.  The AI translators do a fine job.  The OCR wasn’t too bad either, except that Anrich used a strange version of “theta” (θ) where the loop is not closed, so Finereader OCR thinks that is an ampersand (&).  Likewise sigma was sometimes handled as beta.  The high-point was always recognised with an asterisk.  And so on.  The accentuation was a mess, of course; but the machine translators do not seem to care.  My new unicode Greek SPIonic-layout keyboard for Windows 11 worked fine.  But … correcting the OCR became tiresome.  And I found myself wondering why I was bothering.  I never intended to translate everything between the covers of Anrich’s two fat volumes.

Thankfully an academic team has now come along and will do professional work on all the St Nicholas texts.  That is as it should be, and I wish them all the best.  My own humble efforts have made the texts more accessible to everyman, and they never had any purpose beyond that.  If they have spurred renewed interest from scholars, then that is better still.

So… what now?

I was quite impressed with how well the modern Greek translations of St Nicholas material were handled by the AI translators, with a bit of sanity-checking from Google Translate.  I really have almost no translations of patristic material into modern Greek.  Indeed I wonder… now that we can work with modern Greek, it might be interesting to see just what already exists in translation in that language!

The only other text that I have in modern Greek translation is the mass of hardly-edited texts under the name of “Ephraem Graecus”.  I have the Phrantzolas edition of these, thanks to a correspondent.  In fact I find that the ancient/medieval Greek of these is in the elderly TLG disks, which most of us have, so I have access to that too.

I fired up Diogenes, which I use to work with that disk, and picked a text at random.  (In fact it was “Sermo unde magi in Hierosolymam ineunt.”)  I copied some of the text, and ran it through Bard AI.  Here’s the text:

Λόγος ὅτε οἱ μάγοι παρεγένοντο εἰς Ἱεροσόλυμα

 Ὅταν ἀγαθοῦ τινος ὁδοιπόρος τύχῃ συνόδου, χαίρει τὸν πόνον τῆς μακρᾶς ὁδοιπορίας κλεπτόμενος ὁμιλίᾳ· ὡς ῥάβδῳ γὰρ ἐρειδόμενος λόγῳ ἀκονιστικῷ γλώττῃ, συμβαδίζειν κεκονισμένῳ δοκεῖ τῷ ποδὶ καὶ τῷ στόματι ἀκαμάτῳ· μεριζόμενος γόνασι κόπον, κουφίζει χείλεσι πολυβάδιστον βῆμα.  Οὕτω δὴ καὶ τοῦ Χριστοῦ γεννηθέντος, οἱ μάγοι τὸν ἀστέρα ἰδόντες καὶ τοῦτον λαβόντες συνοδοιπόρον, τὸν πολυπόρευτον κόπον ἔκλεπτον τῆς ὁδείας ἐρωτήσει κοπούμενοι, ποῦ ἐστιν ὁ βασιλεὺς πυθόμενοι· ὡς κλέπτας τοῦ τεχθέντος ἠρεύνουν φωνῇ τοὺς Ἑβραίους.  Τοῖς δὲ ἐρωτῶσιν εἰκὸς Ἰουδαῖοι, τί δή, ξένοι, τολμᾶτε, τί φατε, ἄνδρες, φασί; Τί φέροντες ἐπικίνδυνον ἥκατε φήμην; Τί βασιλέα καινὸν βασιλευομένῃ σαλπίζετε πόλει; Τί πρὸς ἄωρον κυβιστεύετε τέλος; Τί κατ’ οἰκείων μαχαιροῦτε γλῶτταν τραχήλων; Τί τάφον ἐπιφέρεσθε στόματι, καθεύδοντα καθ’ ἑαυτῶν διυπνίζετε θάνατον; Ἠπόρει μνημάτων Περσίς, ἵνα ἔτι ζῶντος Ἡρῴδου βασιλέως ἄλλου πυνθάνεσθε; Πολλὴν ἀκούσας ὁμολογήσει χάριν ὑμῖν καὶ μεγάλοις ὑμᾶς ἀμείψειε δώροις.  Ἀλλ’ ἡ πρὸς ταῦτα τοῖς μάγοις ἀπόκρισις σύντομος· εἴδομεν αὐτοῦ, φασί, τὸν ἀστέρα ἐν τῇ ἀνατολῇ καὶ ἤλθομεν προσκυνῆσαι αὐτῷ. Οὐκ ἠρκέσθησαν ἐρωτῆσαι, ἀλλὰ καὶ προσκύνησιν φῆσαι· διὰ τῆσδε τῆς λέξεως Θεὸν εἶναι τὸν τεχθέντα δηλοῦντες.  Ὡς δὲ ἦλθε τῶν τοιούτων πρὸς Ἡρῴδην ἡ φήμη, τοὺς μάγους τέως ἀνείς, τοὺς σοφοὺς Ἰουδαίων καλεῖ λέγων· ποῦ ὁ Χριστὸς γεννᾶται; Οἱ δέ φασιν, ἐν Βηθλεὲμ τῆς Ἰουδαίας.  Ὢ τοῦ θαύματος! Τὸν τόπον ἐπίστανται, καὶ τὸν τόκον βδελύττονται· τὴν Βηθλεὲμ βοῶσι, καὶ τὴν οἰκονομίαν σιγῶσι· τὴν Ἰουδαίαν σημειοῦνται, καὶ τὴν δεσποτείαν ἀρνοῦνται. Εἰ οὕτω γέγραπται, διατί μὴ συντίθεσθε; Εἰ ἀνέγνωτε, διατί μὴ πιστεύετε; Εἰ ἐν Βηθλεὲμ τεχθήσεται, διατί μὴ παρ’ ὑμῶν προσκυνηθήσεται; Ὅθεν συνιεὶς ὁ Ἡρῴδης ὡς οὐχ ἁπλῶς ἥκεσαν ἐρωτῶντες οἱ μάγοι, λάθρα τούτους καλέσας ἐπυνθάνετο παρ’ αὐτῶν τὸν χρόνον τοῦ φαινομένου ἀστέρος, ἀκριβῶς, λέγων, τὸν τρόπον ὃν ἥκατε φάσκοντες, εἴπατε τὸ τῆς ὑμετέρας <ὁδείας>

And here is the output from Bard AI, unedited:

The Story of the Magi

When a traveler meets a good companion, he rejoices, stealing away the pain of the long journey with conversation. For as if leaning on a staff with a sharp-pointed tongue, he seems to walk with ease, his foot and his mouth tireless. Dividing the fatigue on his knees, he lightens his weary step with his lips.

So too, when Christ was born, the Magi, seeing the star and taking it as their companion, stole away the long journey of the road, inquiring and asking, “Where is the king?” They sought out the voice of the Hebrews, as thieves of the newborn.

It is likely that the Jews asked the questioners, “What do you mean, strangers? What are you saying, men? What dangerous rumor do you bring? Why do you trumpet a new king in a city that is already ruled? Why do you stir up trouble at an inopportune time? Why do you sharpen your tongues against your own necks? Why do you bring a tomb to your lips, awakening death against yourselves? Persia is full of tombs. Why do you ask about another king while Herod is still alive? He would hear your confession with great gratitude and reward you with great gifts.”

But the Magi’s answer to this was short: “We saw his star in the east, and we have come to worship him.” They were not satisfied with asking, but they also said that they would worship him. By this word they declared that the one who was born was God.

When word of this came to Herod, he dismissed the Magi for a time and called the wise men of the Jews, saying, “Where is the Christ born?” And they said, “In Bethlehem of Judea.”

Oh, the wonder! They know the place, and they abhor the birth. They cry out Bethlehem, and they are silent about the economy. They point to Judea, and they deny the lordship. If it is so written, why do you not agree? If you have read, why do you not believe? If he is to be born in Bethlehem, why will he not be worshiped by you? Therefore, Herod, realizing that the Magi had not come simply to ask questions, called them secretly and inquired of them the time of the appearance of the star, saying, “Tell me exactly the way you came, and tell me the time of your journey.”

And here, for comparison, is the Google Translate text, again unedited:

Reason when the magicians were brought to Jerusalem

It was a fellowship of the Synod, he enjoys the pain of the long -standing manner of the mackerel; ITON VIMA.   Even so, when Christ was born, the magicians, seeing the star and receiving this as a companion, asked the rich laborer, the thief of the covenant, when they were cut down, where the king was; they sought out the Jews by voice as the thief of what had been done.   And they will ask them if they are Jews, what do you say, foreigners, you dare, what do you eat, men, fasi?  What dangerous fame have you come? What city do you trumpet the king’s new reign? What do you end up turning against the hour? Why do you stab the throats of necks? Persis, the treasurer of monuments, why should you marry Herod’s living king? He heard many confessions for your sake and rewarded you with great gifts.  But the answer to this to the magicians was short: we saw the star in the east, and we came to worship him. It was not enough to ask, but also to worship; through this word God is declaring what has been done.   And when the rumor of these came to Herod, he called the magicians of old, the wise men of the Jews, saying: Where is Christ born? They said, in Bethlehem of Judea.   Ὢ of the miracle! The place they visit, and the interest they detest; Bethlehem loudly, and the economy quietly; Judea they mark, and despotia they deny. If it is written like this, why do you not compose it? If you read it, why do you not believe? If it is made in Bethlehem, why does it not worship before you? How did Herod know that the magicians did not simply come inquiring, and you secretly called them a priest for them at the time of the vision? stay star, exactly, saying, the way you came, you said that of the mother <Odeia>

Google Translate is now out-dated, but because it uses a different technology – NMT – it acts as a useful check on AI.  For instance the first sentence is paraphrased by AI, rather than translated.  At least one can count the sentences and get an idea if it’s all there!

Likewise Diogenes allows you to click on individual words and get the L&S result for each, which  helps in checking.

But all the same, the AI translation looks wonderful.  Basically we can now make use of it for ancient and medieval Greek.  So long as we proceed with caution!

I’m not sure whether I want to start working on Ephraem Graecus tho.  What is there in this mass of texts that is going to be interesting?  At the moment I don’t know.

There is another issue with the Ephraem Graecus material.  The edition was made by Assemani, in the 18th century.  He just printed in a heap whatever he found in the manuscripts.  What this means is that some short “texts” are actually just abstracts of other texts.

But which ones?  There ought to be a list, but if so, it has not reached me.

I wonder whether we could get AI to work out the relationships?  After all, the task is basically one of text comparison.  We have all the Greek in electronic form, thanks to the wonder that is the TLG.  So… can we get AI to look through it and tell us?

I think it might be possible.  But there’s only one way to find out, which is to try.  When I get a break, I might experiment a bit!

Share

How does “AI translation” work? Some high-level thoughts

The computer world is a high-bullshit industry.   Every computer system consists of nothing more than silicon chips running streams of ones (1) and zeros (0), however grandly this may be dressed-up.  The unwary blindly accept and repeat the words and pictures offered by salesmen with something to sell.  These are repeated by journalists who need something to write about.  Indeed the IT industry is the victim of repeated fads.  These are always hugely oversold, and they come, reach a crescendo, and then wither away.  But anybody doing serious work needs to understand what is going on under the hood.  If you cannot express it in your own words, you don’t understand it, and you will make bad decisions.

“AI” is the latest nonsense term being pumped by the media.  “Are the machines going to take over?!” scream the journalists.  “Your system needs AI,” murmur the salesmen.  It’s all bunk, marketing fluff for the less majestic-sounding “large language models (LLM) with a chatbot on the front.”

This area is the preserve of computer science people, who are often a bit strange, and are always rather mathematical.  But it would seem useful to share my current understanding as to what is going on, culled from a number of articles online.   I guarantee none of this; this is just what I have read.

Ever since Google Translate, machine translation is done by having a large volume of texts in, say, Latin, a similarly large volume in English, and a large amount of human-written translations of Latin into English.  The “translator” takes a Latin sentence input by a human, searches for a text containing those words in the mass of Latin texts, looks up the existing English translation of the same text, and spits back the corresponding English sentence.  Of course they don’t just have sentences; they have words, and clauses, all indexed in the same way.  There is much more to this, particularly in how material from one language is mapped to material in the other, but that’s the basic principle.  This was known as – jargon alert – “Neural Machine Translation” (NMT).

This process, using existing translations, is why the English translations produced by Google Translate would sometimes drop into Jacobean English for a sentence, or part of it.

The “AI translation” done using an LLM is a further step along the same road, but with added bullshit at each stage.  The jargon word for this technology seems to be “Generative AI”.

A “large language model” (LLM) is a file.  You can download them from GitHub.  It is a file containing numbers, one after another.  Each number represents a word, or part of a word.  The numbers are not random either – they are carefully crafted and generated to tell you how that word fits into the language.  Words relating to similar subjects have numbers which are “closer together”.  So in a sentence “John went skiing in the snow,” both “snow” and “skiing” relate to the same subject, and will have numbers closer together than the same number for “John.”

Again you need a very large amount of text in that language on both sides.  For each language, these texts are then processed into this mass of numbers.  The numbers tell you whether the word is a verb or a noun, or is a name, or is often found with these words, or never found with those.  The mass of numbers is a “language model”, because it contains vast amounts of information about how the language actually works.  The same English word may have more than one number; “right” in “that’s right” is a different concept to the one in “the politicians of the right.”  The more text you have, the more you can analyse, and the better your model of the language will be.  How many sentences contain both “ski” and “snow”?  And so on.  The model of how words, sentences, and so on are actually used, in real language texts, becomes better, the more data you put in.  The analysis of the texts starts with human-written code that generates connections; but as you continue to process the data, the process will generate yet more connections.

The end result is these models, which describe the use of the language.  You also end up with a mass of data connecting the two together.  The same number in one side of the language pair will also appear in the other model, pointing to the equivalent word or concept.  So 11050 may mean “love” in English but “am-” in Latin.

As before, there are a lot of steps to this process, which I have jumped over.  Nor is it just a matter of individual words; far from it.

The term used by the AI salesmen for this process is “training the model.”  They use this word to mislead, because it gives to the reader the false impression of a man being trained.  I prefer to say “populating” the model, because it’s just storing numbers in a file.

When we enter a piece of Latin text in an AI Translator, this is encoded in the same way.  The AI system works out what the appropriate number for each token – word or part-word – in our text is.  This takes quite a bit of time, which is why AI systems hesitate on-screen.  The resulting stream of encoded numbers are then fed into the LLM, which sends back the corresponding English text for those numbers, or numbers which are mathematically “similar”.  Plus a lot of tweaking, no doubt.

But here’s the interesting bit.  The piece of Latin that we put in, and the analysis of it, is not discarded.  This is more raw data for the model.  It is stored in the model itself.

This has two interesting consequences.

The first consequence is that running the same piece of text through the LLM twice will always give different results, and not necessarily better ones.  Because you can never run the same text through the same LLM twice; the LLM is different now, changed to include your text.

The second consequence is even more interesting: you can poison a model by feeding it malicious data, designed to make it give wrong results.  It’s all data, at the end of the day.  The model is just a file.  It doesn’t know anything.  All it is doing is generating the next word, dumbly.  And what happens if the input is itself AI-generated, but is wrong?

In order to create a model of the language and how it is used, you need as much data as possible.  Long ago Google digitised all the books in the world, and turned them into searchable text, even though 80% of them are in copyright.  Google Books merely gives a window on this database.

AI providers need lots of data.  But one reason why they have tried to conceal what they are doing is, in part, because the data input is nearly all in copyright.  One incautious AI provider did list the sources for its data in an article, and these included a massive pirate archive of books.   But they had to get their data from somewhere.  Similarly this is why there are free tiers to all the AI websites – they want your input.

So… there is no magic.  There is no sinister machine intelligence sitting there.  There is a file full of numbers, and processes.

The output is not perfect.  Even Google Translate could do some odd things.  But AI Translate can produce random results – “hallucinations”.

Further reading

Share

Methodius ad Theodorum (BHG 1352y) Part 4 – A Draft Translation using AI

Sometimes the only way forward is to plunge in, and see what happens.  So I have taken the modern Greek translation of Methodius ad Theodorum by Ch. Stergioulis, and machine-translated it into English.  The results are attached, together with Stergioulis original, which has the ancient Greek facing the modern Greek, and footnotes at the end.

The ancient Greek original is preserved in a single Vatican manuscript, which gave the editor, G. Anrich, a lot of problems – so much so that he printed it in volume 1 of his Agios Nikolaos, and then printed a transcription of the manuscript in vol. 2, with corrections.  Some of his footnotes in vol. 1 betray bafflement; and so do some of Stergioulis’ footnotes!  Words otherwise unknown, I gather.  It’s clear that the text is corrupt.

I scanned Stergioulis’ translation into a Word document (attached).  Then for each chapter I did the following:

  1. Run the Greek through Google Translate and paste it into a Word document.
  2. Send a request to Bard AI, “please translate this from modern Greek into English, with notes”, followed by the Greek.  The notes were usually useless, but sometimes not.
  3. Send a request to ChatGPT 3.5, “please translate this from modern Greek into English” (this wouldn’t give notes), followed by the Greek.
  4. Manually modify the Google Translate output in the Word document using the output from the two AI websites.  Where all three agreed, this was no problem, and it was just a case of choosing the most pleasant version.  Where one disagreed, I looked a bit harder at it.  Where all three disagreed, I started looking up some key words in Greek dictionaries, and looked at the footnotes.  There was  in fact only one sentence, in chapter 10, where the meaning of a clause -“και φοβούμενος το λιχνιστήρι34 αυτής της ανθρωπαρέσκειας για την απόκτηση αγαθών – was completely unclear, and I got there in the end.

This was usually straightforward, not least because the chapters were small.

For a couple of chapters Bard AI threw a wobbly.  Instead of outputting a translation, it started to spew a message in Greek, basically saying “I am an AI model, I don’t know how to do that.”  It seems that it was ignoring the English part of my request string, thought that I was writing in Greek, and so it treated the Greek passage as itself a request to do something, not as something to translate.  I found that replying with “Could you translate that request from modern Greek into English?” did the trick, and caused it to translate instead.

Anyway, here is the output from this process.

The translation is readable enough.  I’m not sure how accurate it is – any comments, anyone?

UPDATE: final version here.

Share

The perils of AI translation

Rather excited by the discoveries that AI would translate medieval Greek, I thought I’d try another attempt at that Ge`ez text that I put into Google Translate some time back.  That is a homily on St Garima by a certain bishop John.  I found the text on my disk, and put a paragraph into Bard AI.  Nope.  It wouldn’t play.  Then I tried ChatGPT 3.5.  That churned out the Nicene Creed, as a supposed translation.

You can’t trust AI.  It can and will generate garbage.  You have to be able to check.

Share

Methodius ad Theodorum (BHG 1352y) using AI on medieval Greek – part 3

In the last post, we tried out various AI tools to translate the modern Greek version of Methodius ad Theodorum into English.  But in the previous post commenter Diego had considerable success doing the same with the original medieval Greek of chapter 3!  So I thought it might be interesting to see what might be done with the medieval Greek of chapter 4.

So I scanned the text using Abbyy Finereader Pro 15 – which treated it as modern Greek, and mangled the accents – and then ran it through (A) Bard AI, (B) ChatGPT 3.5, (C) Google Translate.  I had to split the sentences somewhat, and this wasn’t preserved in the AI translations, so I manually adjusted it.

Here’s the output:

4. Ἀλλά τούτο, εί καὶ παρέλκον, πλήν εις έγκεντρισμον θείου φόβου διηγήσασθαι συνεώρακα, ϊνα μνησθώμεν καὶ φοβηθέντες έκφύγωμεν τήν λέγουσαν απειλήν
A: But even though this is a digression, I have considered telling it for the sake of instilling the fear of God, so that we may remember and, being afraid, flee the threat that says,
B: “However, I have observed that even though it may be enticing, except for the purpose of instilling a divine fear, I narrate it simultaneously, so that we may remember and, being afraid, escape the threatening voice that says,
C: 4. But this, too, is irrelevant, except in the concentration of divine fear, as you tell us, we pray and, being afraid, flee from the so-called threat

«οὐ μὴ καταμείνη τό πνεύμά μου έν τοϊς άνθρώποις τούτοις διά τὸ είναι αυτούς σάρκας».
A: “My Spirit will not remain in these people because they are flesh.”
B: ‘My spirit will not remain among these people because they are flesh.’
C: “Do not let my spirit dwell in these people because they are flesh”.

είτα ευθύς ή σμικρότης τών ήμερων έπετιμήθη· φησίν γάρ·
A: Then immediately the shortness of the days was rebuked; for it says,
B: Then, immediately, the brevity of their days was reproved; for it says,
C: Behold the straightness or the smallness of the meek is despised;

«έσονται δέ αἱ ήμέραι αύτών ἑκατὸν είκοσι έτη»
A: “Their days will be one hundred and twenty years,”
B: ‘Their days will be one hundred and twenty years,’
C: “And the days of these are a hundred and twenty years”

, και πάλιν·
A: and again,
B: and again,
C: , and again;

«κραυγή Σοδόμων και Γομόρρας άνέβη πρός με· καταβάς οϋν δψομαι, εί κατά τήν κραυγήν αύτών συντελοϋνται· εἰ δὲ μή, ‘ίνα γνω».
A: “The cry of Sodom and Gomorrah has ascended to me; therefore I will go down and see whether they are acting according to their cry; if not, I will know.”
B: ‘The outcry of Sodom and Gomorrah has come up to me. I will go down and see if what they have done is as bad as the outcry that has reached me; if not, I will know.’
C: “the cry of Sodom and Gomorrah went up to me; if I go down, I see, according to their cry they are done; but if not, let me know.”

και γάρ έξ άρχής σαρκομανία καὶ τον παγκόσμιον κατακλυσμόν καὶ τόν Πενταπόλεως έμπρησμὸν έφειλκύσατο, ο’ια καὶ νΰν τήν διηγηθεϊσαν θεηλασίαν.
A: For indeed from the beginning, carnality has brought about the worldwide flood and the burning of Pentapolis, and so now the theophany that has been described.
B: For from the beginning, he drew upon himself carnal indulgence, the worldwide deluge, and the burning of the Pentapolis, just as now he has brought upon himself the divine wrath that I have narrated.
C: and since from the beginning sarcomania and the universal cataclysm and the Pentapolis fire were spared, which they did not tell the story of.

δι’ ὸ έν άγιασμώ τό εαυτού σκεύος έκαστος κτησώμεθα· ούκ έσμέν γάρ εαυτών, ώς ό παρακλήτωρ θεού καὶ ύπερπρεσβενων Χριστού Παύλος ό ύιβηλός κέκραγεν, έπειδή τιμῆς ήγοράσΟημεν.
A: Therefore, let each of us acquire his own vessel in sanctification; for we are not our own, as the Paraclete of God and the ambassador of Christ, Paul the blessed, has cried out, because we have been bought with a price.
B: Therefore, let each one possess his vessel in sanctification, for we are not our own, as the advocate of God and the super-apostle Christ, Paul, cried out, ‘You were bought at a price.’
C: by which I sanctify myself, each one of us is a vessel; we are not for ourselves, as the intercessor of God and the ambassador of Christ, Paul the high priest, because of the honor of our king.

καὶ διά τούτο όφείλομεν δοξάσαι τόν θεόν έν τω σώαατι ήμών· ναός γάρ τού έν ήμϊν άγιου πνεύματός έστιν, ου άπό τον θεού έχομεν.
A: And for this reason we owe it to praise God in our body; for it is a temple of the Holy Spirit that is in us, which we have from God.
B: And for this reason, we must glorify God in our bodies, for our bodies are the temple of the Holy Spirit within us, which we have from God.
C: and for this reason we ought to glorify God in our house; for we are a temple of the Holy Spirit, which we have from God.

καὶ άνειμι λοιπόν τή διηγήσει μετά τὴν ύπόμνησιν καθ’ ειρμόν βαδιούμενος.
A: And I return, then, to the narrative, walking in order after the reminder.
B: Thus, having completed the narration, I proceed with the exhortation according to custom.”
C: and Anemi, therefore, recounted it after the reminder as he walked.

The value of this test is again distorted because so much of it is from the New Testament.

We’re getting something, but the quality isn’t great.  We got better from the modern Greek.  The inferiority of Google Translate (C) is noticeable.

I’ll try chapter 5, and see how we get on.

UPDATE (20 Jan 2024):  I tried combining this with the material from modern Greek.  This involved an awful lot of manual fiddling, because sentences end in different places.  I combined them using the Linux “paste” command, and prefixed each line with something identifiable.  Here are the results.  But I think this is just too time-consuming.

Original : 4. Ἀλλά τούτο, εί καὶ παρέλκον, πλήν εις έγκεντρισμον θείου φόβου διηγήσασθαι συνεώρακα, ϊνα μνησθώμεν καὶ φοβηθέντες έκφύγωμεν τήν λέγουσαν απειλήν
A (Bard AI) : But even though this is a digression, I have considered telling it for the sake of instilling the fear of God, so that we may remember and, being afraid, flee the threat that says,
B (ChatGPT3.5): “However, I have observed that even though it may be enticing, except for the purpose of instilling a divine fear, I narrate it simultaneously, so that we may remember and, being afraid, escape the threatening voice that says,
C (Google Tr) : 4. But this, too, is irrelevant, except in the concentration of divine fear, as you tell us, we pray and, being afraid, flee from the so-called threat
D (Modern Gk) : Διηγήθηκα, λοιπόν, σύντομα τούτο το γεγονός, αν και ήταν περιττό, προκειμένου να μπολιάσω στις ψυχές το θείο φόβο, για να θυμούμαστε με φόβο την απειλή (εν. του θεού) πως:
E (ChatGPT3.5): So, I recounted this event briefly, even though it was unnecessary, in order to instill divine fear into the souls, so that we may remember with fear the threat (of God) that:
F (Bing AI) : So I briefly narrated this event, although it was unnecessary, in order to instill divine fear in our souls, so that we may remember the threat (of God) that
G (ChatGPT4) : I have thus briefly recounted this event, though it was unnecessary, in order to instill divine fear into our souls. We must remember with fear the divine threat that

Original : «οὐ μὴ καταμείνη τό πνεύμά μου έν τοϊς άνθρώποις τούτοις διά τὸ είναι αυτούς σάρκας».
A (Bard AI) : “My Spirit will not remain in these people because they are flesh.”
B (ChatGPT3.5): ‘My spirit will not remain among these people because they are flesh.’
C (Google Tr) : “Do not let my spirit dwell in these people because they are flesh”.
D (Modern Gk) : “δεν θα παραμείνει πλέον το πνεύμα μου στους ανθρώπους τούτους, διότι κυριαρχούνται εξ ολοκλήρου από σαρκικά φρονήματα”17 και έτσι να γλιτώσουμε απ’ αυτήν.
E (ChatGPT3.5): ‘My spirit will no longer remain among these people, for they are entirely dominated by carnal thoughts’17 and thus to be saved from it.
F (Bing AI) : “my spirit will no longer remain with these people, because they are entirely dominated by carnal thoughts”17 and thus save ourselves from it.
G (ChatGPT4) : “my spirit shall no longer remain in these humans, for they are entirely dominated by carnal thoughts” (17) and thus save ourselves from it.

Original : είτα ευθύς ή σμικρότης τών ήμερων έπετιμήθη· φησίν γάρ·
A (Bard AI) : Then immediately the shortness of the days was rebuked; for it says,
B (ChatGPT3.5): Then, immediately, the brevity of their days was reproved; for it says,
C (Google Tr) : Behold the straightness or the smallness of the meek is despised;
D (Modern Gk) : Κι ευθύς αμέσως ορίστηκε ένα μικρό χρονικό διάστημα πριν την καταστροφή. Είπε, δηλαδή:
E (ChatGPT3.5): And immediately, a short period before the destruction was appointed. He said, namely:
F (Bing AI) : And immediately a small period of time was set before the destruction. He said, that is:
G (ChatGPT4) : A short period was immediately set before the destruction. It was stated:

Original : «έσονται δέ αἱ ήμέραι αύτών ἑκατὸν είκοσι έτη»
A (Bard AI) : “Their days will be one hundred and twenty years,”
B (ChatGPT3.5): ‘Their days will be one hundred and twenty years,’
C (Google Tr) : “And the days of these are a hundred and twenty years”
D (Modern Gk) : όλες οι υπολειπόμενες μέρες της ζωής τους θα περιορισθούν μόνο σε εκατόν είκοσι ετη”18.
E (ChatGPT3.5): ‘all the remaining days of their lives will be limited to only one hundred twenty years’18.
F (Bing AI) : all the remaining days of their lives will be limited to only one hundred and twenty years”18.
G (ChatGPT4) : “All their remaining days shall be limited to only one hundred and twenty years” (18).

Original : , και πάλιν·
A (Bard AI) : and again,
B (ChatGPT3.5): and again,
C (Google Tr) : , and again;
D (Modern Gk) : Και συμπλήρωσε:
E (ChatGPT3.5): And he added:
F (Bing AI) : And he added:
G (ChatGPT4) : And it was added:

Original : «κραυγή Σοδόμων και Γομόρρας άνέβη πρός με· καταβάς οϋν δψομαι, εί κατά τήν κραυγήν αύτών συντελοϋνται·
A (Bard AI) : “The cry of Sodom and Gomorrah has ascended to me; therefore I will go down and see whether they are acting according to their cry;
B (ChatGPT3.5): ‘The outcry of Sodom and Gomorrah has come up to me. I will go down and see if what they have done is as bad as the outcry that has reached me;
C (Google Tr) : “the cry of Sodom and Gomorrah went up to me; if I go down, I see, according to their cry they are done;
D (Modern Gk) : “κραυγές πολλές ανέρχονται από τα Σοδομα και τη Γομόρρα προς εμέ• οι αμαρτίες τους είναι πάρα πολύ μεγάλες. Θα κατεβώ, λοιπόν, εκεί, για να δω, αν πράγματι οι αμαρτίες τους είναι όπως οι κραυγές που ανέρχονται προς εμέ ή όχι.
E (ChatGPT3.5): ‘Many cries rise from Sodom and Gomorrah to me; their sins are very great. So I will go down there to see if their sins are as the cries that rise to me or not.
F (Bing AI) : “Many cries are coming from Sodom and Gomorrah to me; their sins are very great. So I will go down there to see if their sins are like the cries that are coming to me or not.
G (ChatGPT4) : “Many cries rise up from Sodom and Gomorrah to me; their sins are exceedingly great. I will go down there, then, to see if their sins are as the cries that rise up to me or not.

Original : εἰ δὲ μή, ‘ίνα γνω».
A (Bard AI) : if not, I will know.”
B (ChatGPT3.5): if not, I will know.’
C (Google Tr) : but if not, let me know.”
D (Modern Gk) : Οπωσδήποτε θέλω να μάθω!”19.
E (ChatGPT3.5): I definitely want to know!’19.
F (Bing AI) : I definitely want to know!”19.
G (ChatGPT4) : I definitely want to know!” (19).

Original : και γάρ έξ άρχής σαρκομανία καὶ τον παγκόσμιον κατακλυσμόν καὶ τόν Πενταπόλεως έμπρησμὸν έφειλκύσατο, ο’ια καὶ νΰν τήν διηγηθεϊσαν θεηλασίαν.
A (Bard AI) : For indeed from the beginning, carnality has brought about the worldwide flood and the burning of Pentapolis, and so now the theophany that has been described.
B (ChatGPT3.5): For from the beginning, he drew upon himself carnal indulgence, the worldwide deluge, and the burning of the Pentapolis, just as now he has brought upon himself the divine wrath that I have narrated.
C (Google Tr) : and since from the beginning sarcomania and the universal cataclysm and the Pentapolis fire were spared, which they did not tell the story of.
D (Modern Gk) : Πράγματι, εξ αρχής η σαρκολατρεία20 προκάλεσε και επέφερε τον παγκόσμιο κατακλυσμό και την πυρπόληση της Πεντάπολης21, όπως και τη μόλις διηγηθείσα θεϊκή τιμωρία (των Πατάρων).
E (ChatGPT3.5): Indeed, from the beginning, idolatry20 provoked and brought about the worldwide flood and the destruction of Pentapolis21, as well as the divine punishment (of the Patari).
F (Bing AI) : Indeed, from the beginning, idolatry20 caused and brought about the worldwide flood and the burning of Pentapolis21, as well as the divine punishment (of the Patari).
G (ChatGPT4) : Indeed, from the beginning, flesh worship (20) caused and brought about the global flood and the burning of the Pentapolis (21), as well as the divine punishment of the Patara just narrated.

Original : δι’ ὸ έν άγιασμώ τό εαυτού σκεύος έκαστος κτησώμεθα·
A (Bard AI) : Therefore, let each of us acquire his own vessel in sanctification;
B (ChatGPT3.5): Therefore, let each one possess his vessel in sanctification,
C (Google Tr) : by which I sanctify myself, each one of us is a vessel;
D (Modern Gk) : Για το λόγο αυτό ας διαφυλάττει ο καθένας το σώμα του ως σκεύος αγιασμού’2.
E (ChatGPT3.5): For this reason, let everyone preserve his body as a vessel of sanctification.
F (Bing AI) : For this reason, let everyone preserve his body as a vessel of sanctification’2.
G (ChatGPT4) : For this reason, let each one keep his body as a vessel of sanctification (22).

Original : ούκ έσμέν γάρ εαυτών, ώς ό παρακλήτωρ θεού καὶ ύπερπρεσβενων Χριστού Παύλος ό ύιβηλός κέκραγεν, έπειδή τιμῆς ήγοράσΟημεν.
A (Bard AI) : for we are not our own, as the Paraclete of God and the ambassador of Christ, Paul the blessed, has cried out, because we have been bought with a price.
B (ChatGPT3.5): for we are not our own, as the advocate of God and the super-apostle Christ, Paul, cried out, ‘You were bought at a price.’
C (Google Tr) : we are not for ourselves, as the intercessor of God and the ambassador of Christ, Paul the high priest, because of the honor of our king.
D (Modern Gk) : Δεν ανήκουμε στον εαυτό μας, όπως έντονα διακήρυττε ο μέγας Παύλος, αυτός που διαρκώς ικέτευε τον θεό και παρακαλούσε συνεχώς τον Χριστό, γιατί έχουμε εξαγορασθεί με πολύτιμο τίμημα, το αίμα του Χριστού.
E (ChatGPT3.5): We do not belong to ourselves, as the great Paul strongly proclaimed, he who constantly implored God and continually besought Christ because we have been redeemed with a precious ransom, the blood of Christ.
F (Bing AI) : We do not belong to ourselves, as the great Paul strongly declared, he who constantly implored God and continuously prayed to Christ, because we have been redeemed with a precious price, the blood of Christ.
G (ChatGPT4) : We do not belong to ourselves, as the great Paul fervently proclaimed, who constantly supplicated God and continuously implored Christ, for we have been bought with a precious price, the blood of Christ.

Original : καὶ διά τούτο όφείλομεν δοξάσαι τόν θεόν έν τω σώαατι ήμών· ναός γάρ τού έν ήμϊν άγιου πνεύματός έστιν, ου άπό τον θεού έχομεν.
A (Bard AI) : And for this reason we owe it to praise God in our body; for it is a temple of the Holy Spirit that is in us, which we have from God.
B (ChatGPT3.5): And for this reason, we must glorify God in our bodies, for our bodies are the temple of the Holy Spirit within us, which we have from God.
C (Google Tr) : and for this reason we ought to glorify God in our house; for we are a temple of the Holy Spirit, which we have from God.
D (Modern Gk) : Γι’ αυτό και οφείλουμε να δοξάσουμε τον θεό με το σώμα μας, χωρίς να το μολύνουμε• γιατί είναι ναός του Αγίου Πνεύματος που κατοικεί μέσα μας, και το οποίο έχουμε λάβει από τον θεό23.
E (ChatGPT3.5): Therefore, we must glorify God with our bodies without defiling them, for it is the temple of the Holy Spirit that dwells within us, which we have received from God23.
F (Bing AI) : Therefore, we must glorify God with our body, without defiling it; because it is the temple of the Holy Spirit that dwells within us, and which we have received from God23.
G (ChatGPT4) : Therefore, we owe it to glorify God with our body, without defiling it; for it is the temple of the Holy Spirit dwelling within us, which we have received from God (23).

Original : καὶ άνειμι λοιπόν τή διηγήσει μετά τὴν ύπόμνησιν καθ’ ειρμόν βαδιούμενος.
A (Bard AI) : And I return, then, to the narrative, walking in order after the reminder.
B (ChatGPT3.5): Thus, having completed the narration, I proceed with the exhortation according to custom.”
C (Google Tr) : and Anemi, therefore, recounted it after the reminder as he walked.
D (Modern Gk) : Προχωρώ, λοιπόν, μετά την υπόμνηση των παραπάνω στη διήγηση, και ολόγος μου θα έχει ως εξής:
E (ChatGPT3.5): I proceed, therefore, after the reminder of the above in the narrative, and my discourse will be as follows:
F (Bing AI) : So, after the above reminder, I proceed with the narrative, and my speech will be as follows:
G (ChatGPT4) : I proceed, therefore, after reminding the above in the narration, and my speech will be as follows:

Hmm.

Share

Methodius ad Theodorum (BHG 1352y) in modern Greek – part 2

Using AI and a dictionary, let’s try out the translation approach from my last post on a further chapter of this modern Greek translation of the Life of St Nicholas by Methodius.  Here’s the text, hopefully with few OCR errors:

4. Διηγήθηκα, λοιπόν, σύντομα τούτο το γεγονός, αν και ήταν περιττό, προκειμένου να μπολιάσω στις ψυχές το θείο φόβο, για να θυμούμαστε με φόβο την απειλή (εν. του θεού) πως: “δεν θα παραμείνει πλέον το πνεύμα μου στους ανθρώπους τούτους, διότι κυριαρχούνται εξ ολοκλήρου από σαρκικά φρονήματα”17 και έτσι να γλιτώσουμε απ’ αυτήν. Κι ευθύς αμέσως ορίστηκε ένα μικρό χρονικό διάστημα πριν την καταστροφή. Είπε, δηλαδή: όλες οι υπολειπόμενες μέρες της ζωής τους θα περιορισθούν μόνο σε εκατόν είκοσι ετη”18. Και συμπλήρωσε: “κραυγές πολλές ανέρχονται από τα Σοδομα και τη Γομόρρα προς εμέ• οι αμαρτίες τους είναι πάρα πολύ μεγάλες. Θα κατεβώ, λοιπόν, εκεί, για να δω, αν πράγματι οι αμαρτίες τους είναι όπως οι κραυγές που ανέρχονται προς εμέ ή όχι. Οπωσδήποτε θέλω να μάθω!”19. Πράγματι, εξ αρχής η σαρκολατρεία20 προκάλεσε και επέφερε τον παγκόσμιο κατακλυσμό και την πυρπόληση της Πεντάπολης21, όπως και τη μόλις διηγηθείσα θεϊκή τιμωρία (των Πατάρων). Για το λόγο αυτό ας διαφυλάττει ο καθένας το σώμα του ως σκεύος αγιασμού’2. Δεν ανήκουμε στον εαυτό μας, όπως έντονα διακήρυττε ο μέγας Παύλος, αυτός που διαρκώς ικέτευε τον θεό και παρακαλούσε συνεχώς τον Χριστό, γιατί έχουμε εξαγορασθεί με πολύτιμο τίμημα, το αίμα του Χριστού. Γι’ αυτό και οφείλουμε να δοξάσουμε τον θεό με το σώμα μας, χωρίς να το μολύνουμε• γιατί είναι ναός του Αγίου Πνεύματος που κατοικεί μέσα μας, και το οποίο έχουμε λάβει από τον θεό23. Προχωρώ, λοιπόν, μετά την υπόμνηση των παραπάνω στη διήγηση, και ολόγος μου θα έχει ως εξής:

Next, the pre-AI Google Translate output:

4. Therefore, I briefly related this event, although it was unnecessary, in order to instill in souls the divine fear, so that we remember with fear the threat (of God) that: “my spirit will no longer remain among men these, because they are completely dominated by carnal thoughts”17 and thus escape from it. And right away a short period of time was set before the disaster. He said, in other words: all the remaining days of their lives will be limited to only one hundred and twenty years.”18 And he added: “many cries go up from Sodom and Gomorrah to me; their sins are very great. I will therefore come down there, to see whether their sins are really like the cries that come up to me or not. I certainly want to know!”19 Indeed, from the beginning flesh worship20 caused and brought about the global flood and the burning of the Pentapolis21, just as the just-narrated divine punishment (of the Patras). For this reason let each one guard his body as a vessel of sanctification’2. We do not belong to ourselves, as the great Paul strongly proclaimed, he who constantly begged God and constantly begged him Christ, for we have been redeemed with a precious price, the blood of Christ. Therefore we ought to glorify God with our body, without defiling it; for it is a temple of the Holy Spirit who dwells in us, and which we have received from the god23. So, after reminding the above, I proceed to the narration, and my thesis will be as follows:

Now time for AI.  I asked “Can you translate something for me from modern Greek into English.”  The ChatGPT 3.5 output using a free account:

The narrator recounted, therefore, briefly this event, although it was unnecessary, in order to instill divine fear into the souls, so that we may remember with fear the threat (of God) that: “My spirit will no longer remain among these people, for they are entirely dominated by carnal thoughts”17 and thus to be saved from it. And immediately a small period before the destruction was appointed. He said, namely: “all the remaining days of their lives will be limited to one hundred twenty years”18. And he added, “Many cries rise from Sodom and Gomorrah to me; their sins are very great. So I will go down there to see if their sins are as the cries that rise to me or not. I definitely want to know!”19. Indeed, from the beginning, idolatry20 provoked and brought about the worldwide flood and the destruction of Pentapolis21, as well as the divine punishment (of the Patari). For this reason, let everyone preserve his body as a vessel of sanctification. We do not belong to ourselves, as the great Paul strongly proclaimed, he who constantly implored God and continually besought Christ, because we have been redeemed with a precious ransom, the blood of Christ. Therefore, we must glorify God with our bodies without defiling them, for it is the temple of the Holy Spirit that dwells within us, which we have received from God23. I proceed, therefore, after the reminder of the above in the narrative, and my discourse will be as follows:

That seemed a bit odd, so I tried again:

So, I recounted this event briefly, even though it was unnecessary, in order to instill divine fear into the souls, so that we may remember with fear the threat (of God) that: ‘My spirit will no longer remain among these people, for they are entirely dominated by carnal thoughts’17 and thus to be saved from it. And immediately, a short period before the destruction was appointed. He said, namely: ‘all the remaining days of their lives will be limited to only one hundred twenty years’18. And he added: ‘Many cries rise from Sodom and Gomorrah to me; their sins are very great. So I will go down there to see if their sins are as the cries that rise to me or not. I definitely want to know!’19. Indeed, from the beginning, idolatry20 provoked and brought about the worldwide flood and the destruction of Pentapolis21, as well as the divine punishment (of the Patari). For this reason, let everyone preserve his body as a vessel of sanctification. We do not belong to ourselves, as the great Paul strongly proclaimed, he who constantly implored God and continually besought Christ because we have been redeemed with a precious ransom, the blood of Christ. Therefore, we must glorify God with our bodies without defiling them, for it is the temple of the Holy Spirit that dwells within us, which we have received from God23. I proceed, therefore, after the reminder of the above in the narrative, and my discourse will be as follows:

Now Bing CoPilot:

  1. 4.  So I briefly narrated this event, although it was unnecessary, in order to instill divine fear in our souls, so that we may remember the threat (of God) that “my spirit will no longer remain with these people, because they are entirely dominated by carnal thoughts”17 and thus save ourselves from it. And immediately a small period of time was set before the destruction. He said, that is: all the remaining days of their lives will be limited to only one hundred and twenty years”18. And he added: “Many cries are coming from Sodom and Gomorrah to me; their sins are very great. So I will go down there to see if their sins are like the cries that are coming to me or not. I definitely want to know!”19. Indeed, from the beginning, idolatry20 caused and brought about the worldwide flood and the burning of Pentapolis21, as well as the divine punishment (of the Patari). For this reason, let everyone preserve his body as a vessel of sanctification’2. We do not belong to ourselves, as the great Paul strongly declared, he who constantly implored God and continuously prayed to Christ, because we have been redeemed with a precious price, the blood of Christ. Therefore, we must glorify God with our body, without defiling it; because it is the temple of the Holy Spirit that dwells within us, and which we have received from God23. So, after the above reminder, I proceed with the narrative, and my speech will be as follows:

Well, this chapter is studded with quotations from scripture, so that skews things a bit.  One would expect scripture to be handled OK by pretty much any system.

Now again there are odd words.  σαρκολατρεία is rendered as “flesh worship” by Google Translate, and “idolatry” by AI.  WordReference.com doesn’t know the word.  Hum.

But what strikes me most here is that I need to compare closely the various different versions of text and translation.  Doing it on a paragraph is cumbersome.  I need to split the text into sentences, and interleave the various versions.

It’s also rather cumbersome using the websites to access ChatGPT.  There is a command-line interface.

I need to upgrade my tools before I do any more on this.

Share

AI Translation of modern Greek once more

In my last two posts here and here, I looked at AI and other translations of the following passage from Ch. Stergioulis’ modern Greek translation of the “Life of St Nicholas” by Methodius, Archbishop of Constantinople (the “Methodius ad Theodorum”:

3. Καταγόταν ο Νικόλαος από τα Πάταρα, πόλη της επαρχίας των Λυκίων, η οποία εκείνη την εποχή (του αγίου Νικολάου) είχε αρκετούς κατοίκους. Τώρα όμως μοιάζει περισσότερο με κώμη παρά με πόλη14. Κι αυτό παθαίνουν πολυτρόπως οι πόλεις-παρακμάζουν και οι κάτοικοί τους μεταναστεύουν-εξαιτίας των αμέτρητων κριμάτων τους έναντι του θεού. Έτσι λοιπόν κατανοούμε και αυτό που συνηθίζουμε να γράφουμε, ότι δηλαδή οι πόλεις τιμωρούνται για τις αμαρτίες των κατοίκων τους15. Βρίσκεται δηλαδή, εκεί κοντά τόπος πεδινός, σχισμένος στα δύο, σαν ξεσχισμένο ιμάτιο. Κι από το ρήγμα που έχει δημιουργηθεί, φαίνεται να αναδύεται, την ημέρα, καυτός ατμός, ενώ τη νύχτα ξεπηδά σαν από χάλκινο καμίνι φλογώδης καπνός16. Και καίει το χέρι όποιου τολμά να το ακουμπήσει, χωρίς όμως να του το κατατρώγει αλλά και την σάρκα, όπως θα γινόταν στην περίπτωση που κάποιος απλώνει το χέρι του στη φωτιά• είναι όμως απεικόνιση και προοίμιο, με μικρή ομοιότητα, του αιωνίου πυρός. Και ευρισκόμενοι σε απορία με το φαινόμενο οι μεγαλύτεροι σε ηλικία μετέφεραν από γενιά σε γενιά, ο κάθε πατέρας στο παιδί του, (όπως μας διασώζεται ο άγραφος πλην αληθινός τούτος λόγος), ότι εξαιτίας της ακολασίας και της σαρκολαγνείας των κατοίκων της η θεία δίκη κατεδίκασε τούτη τη γη στο φρικτό αυτό θέαμα, για να αποτρέψει την επι-στροφή τους στο μέρος αυτό.

Today I went back to the same sites, to see what came out.  First I asked ChatGPT 3.5, “Can you translate something from modern Greek to English”, and when it said yes, pasted in the above text:

Nicholas originated from Patara, a city in the province of Lycia, which at that time (of Saint Nicholas) had several inhabitants. However, now it resembles more of a village than a city. This is a fate suffered in various ways by cities—they decline, and their inhabitants migrate—due to their countless sins against God. Therefore, we understand what we commonly write, that is, cities are punished for the sins of their residents.

Nearby, there is a flat place, split in two, like a torn garment. From the rift that has been created, hot steam seems to emerge during the day, while at night it rises like fiery smoke from a bronze furnace. And it burns the hand of anyone who dares to touch it, without, however, consuming it as well as the flesh, as would happen if someone were to extend their hand into a fire. It is, however, a representation and prelude, with little resemblance, to the eternal fire.

In awe of this phenomenon, the older ones passed down from generation to generation, each father to his child (as this unwritten but true account has been preserved for us), that due to the immodesty and sensuality of its inhabitants, divine justice condemned this land to this dreadful spectacle to prevent their return to this place.

This is different to the output I got two days ago.  Mostly it is very similar, and certainly impressive.  In a couple of places, it isn’t as good.  In a couple of places it’s a bit better.

Next, I tried using Bing CoPilot “with GPT-4”:

3.  Nikolaos originated from Pataras, a city in the province of Lycia, which at that time (of Saint Nicholas) had many inhabitants. Now, however, it looks more like a village than a city. And this is what happens to cities – they decline and their inhabitants emigrate – because of their countless sins against God. Therefore, we understand what we are used to writing, that is, that cities are punished for the sins of their inhabitants. There is a nearby flat place, split in two like a torn garment. And from the rift that has been created, hot steam appears during the day, while at night fiery smoke emerges like from a copper furnace. And it burns the hand of anyone who dares to touch it, but without consuming it or the flesh, as would happen if someone were to put their hand in fire. It is, however, a depiction and a prelude, with little similarity, of the eternal fire. And the older ones, puzzled by the phenomenon, passed down from generation to generation, each father to his child (as this unwritten but true story is preserved), that because of the debauchery and carnality of its inhabitants, divine justice condemned this land to this terrible spectacle, to prevent their return to this place.

This is almost the same as I got in my first post.  But one good feature of Bing AI is that it indicates the websites that it is using:

I imagine that this is concealed by other sites, in case they had to pay royalties.

It does feel creepy that the same input on different days produces different output from the same site.  I know that Google Translate could be like that.  But even so… it feels flaky.

All the same, there’s no doubt that AI is producing better results than Google Translate or DeepL.  I do think that we then need to look up variable words and establish their meaning ourselves.

Doing so could certainly be easier. Remember that I know no modern Greek at all; all of this is a way to use online tools.

One bit that varies is the “flat place” or “low lying place” or “plain”.  Looking at Google Translate, I see this preserves the footnote numbers, which helps us work out whereabout in the Greek text this is.  It’s just after footnote 15:

inhabitants15. In other words, there is a plain nearby, torn in two, like a torn garment.

This must be “τόπος πεδινός”, “topos pedinos”.  Even I know “topos” must be a “place”, if only thanks to Tales from Topographic Oceans.  So “pedinos” is our word of not-quite certain meaning.

Going to Lexilogos and copying the word, and clicking on WordReference gives us this page.

This does indeed actually explain the problem – plains are flat and are found in the lowlands, hence the various different meanings. I hate that this is an online-only tool – what if the web goes down? – but it’s something.

Likewise I get αρκετούς – “several inhabitants”?, and plug this in.  Taking off letters from the end, it asks if I mean αρκετός – I know enough Greek to agree – and tells me that it means enough, ample, plenty, sufficient.  I.e. “lots”, which is what I would expect.  Likewise “διασώζεται” also works, once you remove a few letters to get the base verb, meaning “save, rescue”, i.e. “preserve”.

So maybe this is how we use these tools to make a translation:

  1. Use Google Translate on the modern Greek to get a general idea of what each sentence means, and a few signposts so I can move around the Greek text.
  2. Bang the Greek into ChatGPT to get a better translation.  Maybe do the same with Bing AI, if it’s working.
  3. Compare the three, and look for oddities.
  4. Find the actual modern Greek word for each oddity, look it up in WordReference.com, and see if this can resolve it.  (Remembering, if using Chrome, to force it NOT to translate the page into Engish).

That has the look of a way to do the translation. It’s cumbersome.  Any clues about Greek will help mightily.  But it really requires very little knowledge, and you can produce a respectable translation, I think.

I will try it out with another chapter of Methodius ad Theodorum.

UPDATE (17 Jan 2024): Much more in the comments below.

Share

Is Google now doomed? Wild thoughts and conspiracy theories below!

I have just spent an interesting hour on the PC since my last post.  Those who read it will recall that I posted some modern Greek, and then the Google Translate output for it – good, but by no means perfect.  It then occurred to me to try Microsoft’s Bing AI.  The output from that was marvellous, so I added that to the end of the post.

Then I started pasting more of the modern Greek into Bing AI, chapter by chapter.  This went well until chapter 11, when it crashed.  And crashed and crashed.  When it did return, the output was obviously inferior, using different sources!  So I stopped.  The Bing AI interface was frankly a mess anyway.

Then it occurred to me to try Bard AI, the Google product.  This produced… inferior output.  Not really any better than Google Translate.

Then I tried ChatGPT 3.5, the OpenAI product.  And suddenly I was getting good output again, if subtly different.

Google has dominated the web ever since the late 90s.  But it is basically a search engine company.  It rose to power because it was a markedly better search engine than AltaVista, which it destroyed.  But the founders of Google have long since departed, as has its “don’t be evil” ethos.  For many years now the company has concentrated on squeezing dollars out of the web.

But “AI” is basically search-engine technology.  So the rise of OpenAI is really the arrival of a new generation of search engine; and Google is struggling.

What the OpenAI people have realised is that most people do not, actually, want to “search the web”.  They want information.  So they have reimagined the search engine, added a  layer of algorithms on top, with a chat-bot on top of that.  Their search engine produces information, not lists of results.  These engines can be targeted for developers, or translators, or other specialised markets.

It looks very much as if OpenAI have created the next generation search engine.

If so, it must have become obvious to search-engine people about a year ago.  Google executives must have been quaking in their well-funded boots.  This is the end for Google, as Google was the end for AltaVista… unless.  Unless Google can catch up.

It has been telling how little we really know about how “AI” works.  The hype has been tremendous.  The hard facts have been few.  And no wonder, if it’s now a cut-throat commercial race.

Google may have some trouble here.  It isn’t really a tech company any more.  It’s an advertising company.  Products like Google Docs were invented elsewhere and bought in.  Microsoft went down this route, as IBM did before them.  There is still interesting tech going on, but they aren’t the cutting edge.

Microsoft will be the big beneficiary of this new era, because of its far-sighted investment in the OpenAI company.  Bing AI, I think, must basically be a front-end on OpenAI.

If all this is true, then Google may not be able to fight back.  The link to Microsoft means that they cannot just buy OpenAI, which would have been a possible way forward.  So… maybe Google will enter the long decline experienced by other tech companies before.  Once people do not use their search engine, it’s basically over.

For the conspiracy minded, this may also explain some otherwise perplexing events of the last year.  In particular it may explain the boardroom coup at OpenAI, when the directors sacked the company founder, Sam Altman.  Why the heck would you do that?  It didn’t stick – Microsoft intervened. The directors seem to be just corporate drones, of no special talent.  Microsoft wasn’t going to let them wreck the company that was about to take over the world.  But why would you do that at all?  Why get rid of the guy who will make you rich?

If this was a Hollywood movie, who would be the prime beneficiary of those mysterious shenanigens at OpenAI?  If OpenAI had collapsed, if Altman could no longer direct the company, who would benefit?  It’s an interesting thought.  Did someone bribe the renta-board of OpenAI to crash the company?  Some existing vested interest?  Not necessarily Google, of course, but possibly someone with shares in it?

Likewise it is interesting to see all the lawsuits, and claims of copyright on the databases used, the large language models (LLMs).  Not that the claims are bogus.  If I understand correctly, there is little doubt that many of the modern AI databases are indeed based upon masses of copyright material.  We know this thanks to one company making an incautious publication which listed what they were loading into their database!

But say that the copyright claims probably valid.  Whether that is true or not – and when it will be decided – will be a decision for the US courts.  These are notoriously political, so money and power will no doubt decide.  Again, who would benefit from knobbling the new technology?

In truth I have no idea about all of this.  There must be people less connected than myself, but it might be hard to name them!  But perhaps all this really signals that the revolution is here.

Yet this revolution may not benefit us, the ordinary internet users.

For one thing, OpenAI is the first web technology that you have to pay for.  It’s on subscription.  This is a new, and unwelcome change to the free internet.

Likewise searching the web at least gives you a choice of viewpoints.  If you get your news from OpenAI, you get one viewpoint only.  It’s great news for the would-be censors, who seem to grow more numerous every day.

It will be interesting to see what happens.

Share