Sorry, I know there’s been a glut of AI-related material here, but I was disturbed by this in a different way. Rachel Chason writes for the Washington Post (archived):
SAFO, Mali — Most of the students had never seen their native language in its written form until recently. Now, they were eagerly sounding out the words appearing on the ThinkPad laptops before them, sometimes stumbling as they read a story written entirely in Mali’s most popular language, Bambara.
The twist? The story on their screens had been generated, translated and illustrated using artificial intelligence.
As Mali’s relationship with French — the language of its former colonial ruler, France — has grown more fraught, an effort to use AI to create children’s books in Bambara and other local languages is gaining momentum. With political tensions high between the two countries, Mali’s military government last year replaced French as the country’s “official” language, instead elevating Bambara and 12 other native languages, though French will still be used in government settings and public schools.
That change has meant there is more political will behind efforts like that of RobotsMali, a start-up that has used artificial intelligence to create more than 140 books in Bambara since last year, said Séni Tognine, who works in Mali’s Education Ministry and has been helping RobotsMali create its books. Now, he said, both the government and the people “are engaged in wanting to learn and valorize local languages.” […]
The team, whose work was first reported by Rest of World, eliminates examples that would not be relevant to most kids in Mali, then uses Google Translate — which added Bambara in 2022 and employs AI to improve its translations — to do a first round of translation. Experts like Tognine then correct any mistakes. Another staff member uses a variety of AI image creators to illustrate the stories, ensuring that the characters are relatable to Malian kids, and then turns to ChatGPT to create reading comprehension tests. […]
The vast majority of Africa’s roughly 1,000 languages are not represented on websites, which big generative AI platforms like ChatGPT crawl to help train themselves.
If you ask ChatGPT the most basic questions in Ethiopia’s two most popular languages, Amharic and Tigrinya, for example, it produces a nonsensical jumble of Amharic, Tigrinya and sometimes even other languages, Asmelash Teka Hadgu said. But Hadgu, who created a start-up focused on using machine learning to translate between English and Ethiopian languages, said that specific projects like that of RobotsMali also speak to the potential of artificial intelligence.
“If it is done right,” he said, “the potential in terms of democratizing access to education is enormous.” […]
A previous effort by Mali’s government to introduce Bambara into public schools largely failed because of a lack of funding, teacher training and parental interest for children to learn a language in school other than French, Tognine said.
But he said that in recent years, there has been a growing embrace of the importance of learning to read and write the national languages, which have traditionally been primarily spoken, partly because of the government’s rejection of France and focus on national sovereignty.
“It enriches our cultural and linguistic history,” Bakari Sahogo, another member of the Education Ministry who has been working with RobotsMali, said about the importance of writing in Bambara and other local languages. “And [it] permits us to safeguard and develop our culture.”
[Michael] Leventhal, who worked as a tech entrepreneur in Silicon Valley before moving to Mali a decade ago to teach computer science, said the ultimate goal is to use artificial intelligence to help Mali develop a stronger written tradition of Bambara than currently exists. That could happen, he said, as artificial intelligence systems get access to more language data.
All very interesting, but what bothers me is that there is nowhere even a mention of the issue of using “AI” rather than humans to write these books. I’m sure there is a shortage of professional writers, and they probably see it as more cost-effective to have computers spit out texts than train native speakers to do so, but I can’t believe the books so produced are as good as human-produced ones would be, either artistically or linguistically. I can believe the officials involved didn’t want to discuss (or even think about) that, but isn’t it a reporter’s job to bring up such things? (Thanks, Eric!)
I can’t believe the books so produced are as good as human-produced ones would be, either artistically or linguistically.
I agree. Even Spiegel headlines are more and more often off-kilter – in ways that *could* be just simple grammar or lexical mistakes such as people make, but I suspect they’re un-proofed AI produce.
Example: something like “Der Experte empfiehlt in diesen Fällen zu größerer Vorsicht“. That’s understandable but weird, definitely a German as a Second Language artefact. It should be “Der Experte rät in diesen Fällen zu größerer Vorsicht“, or “Der Experte empfiehlt in diesen Fällen größere Vorsicht“.
If such off-kilter language becomes more common, even Germans could become accustomed to talking to each other as if they were talking to the Mexican maid (mut. mutand.)
This is plastic imitation indigenisation.
However, what really seems to be involved is using Google Translate for first drafts of translation from French, and then getting L1 speakers to make it acceptable. (Fair enough, up to a point,)
The “AI” spin is just journalistic crap combined with the usual “AI” boosterism. Vacuous.
Presumably this is really just a recycled press release from “Robotsmali.” I suspect our ace reporter does not seek out stories about writing in the indigenous languages of Mali all that much.
(I can’t access the article, or the archived version.)
That could happen, he said, as artificial intelligence systems get access to more language data
The “tech entrepreneur” feels that more plagiarism will solve the problem. I am … unsurprised.
I can’t access the article, or the archived version.
Well, hell. What’s the point of archives then? I tried!
RobotsMali are pretty into providing helpful stories for the press, I see:
https://robotsmali.org/en/actualites/presse/
(Though the Washington Post seems to be a relatively large catch for them.)
By coincidence, a Michael Leventhal turns up in this story too:
https://www.prospectmagazine.co.uk/ideas/technology/65689/artificial-intelligence-language-translation-twi-ghana
they probably see it as more cost-effective to have computers spit out texts than train native speakers to do so, but I can’t believe the books so produced are as good as human-produced ones would be, either artistically or linguistically
They will have sold this to the Malian military dictatorship as being a cheaper way of supposedly fulfilling some of its ethnonationalist aspirations. Genuine efforts in that direction (of a kind which could actually provide a lasting benefit to ordinary Malians) would require a real commitment to actual living and breathing human beings. In fairness to the military junta, in the current state of that unhappy country, that would be difficult, and dependent on meaningfully addressing many acute political and economic problems.
This is actually a beautiful example of the way in which Large-Scale Automated Plagiarism Engines and kindred technology are doing actual harm now; the bollocks about existential threats to humanity is conjuror’s patter intended to stop us noticing what they are doing with their hands in front of our eyes.
The misty-eyed blether about empowerment of speakers of minority languages (mostly by enabling them to use Facebook in their Own Languages, it appears, so that they can be fed advertising in their Own Languages) goes nicely with cutting as many actual speakers of those languages as possible out of the process, because of the enormous costs of employing poor people.
The archive works fine for me. Maybe the site was down?
Call it “machine translation” instead of AI. I am skeptical. It took a lot of training data, and then wiggling and adjusting, to get machine translation working decently on large languages. Getting software to write acceptable Bambara, with 1/10,000 or whatever the training data — nope. Nopity nope.
This was Firefox, complaining that a “Secure connection failed.” My phone also refused to look at it. Chrome just keeps telling me that “the connection has been reset.”
I tried turning off my various adblocking and javascript-zapping stuff, but it didn’t help.
Mind you, I suspect that I haven’t missed much …
12 other native languages
Case in point …
There are many more than thirteen indigenous languages in Mali.
Furthermore …
I gather (from Jeffrey Heath’s works) that post-independence Mali basically took over the colonial French names for supposed indigenous languages, which (for example) lumped all the Songhay languages together as a “Songhay language”, and similarly with Dogon etc etc; essentially, they weren’t interested in the details of the linguistic situation and used ethnic labels as if they were the same thing.
Then a particular Dogon language (say) was selected as “Dogon”, and that is the only one even notionally used in schools. (In the case of Dogon, this was not even the most widely used language, Jamsay.) This is on a level with telling English primary school children that their “native language” is Dutch.
I suspect that this particular spiffy initiative is actually almost entirely about Bambara.
If there were ~0 children’s books in Bambara prior to this AI/human collaboration, then this effort is a large step in the right direction. Also, the human-edited stories will become part of the Bambara corpus which will lead to more capable AI-generated Bambara in the future.
This is on a level with telling English primary school children that their “native language” is Dutch.
In the case of Bangime, like telling Basque schoolchildren that their native language is Italian.
If there were ~0 children’s books in Bambara prior to this AI/human collaboration
I’m pretty sure that’s not the case.
In looking at the WP article for Bambara to see if there was an estimate for the number of speakers, I discover this excellent reliable source:
https://josepheidelberg.wordpress.com/blog/
Worthy of the Sino-Platonic Papers!
Eidelberg’s WP page (he has one) appears to be a plot to undermine the credibility of Wikipedia. For sheer mindless credulity, unchallenged by any objective editing, only Noam Chomsky’s WP page comes close …
Confess it, you love these people…
Back in 1977, you could believe anything. Elephants with hooves, what have you.
(WP = Wikipedia? WordPress?)
Confess it, you love these people…
Nah. I’m not into freak shows. And exasperation keeps interfering with the compassion I ought to be feeling for them …
Still, If others had not been foolish, we should be so.
https://poets.org/poem/proverbs-hell
[WikiPedia. ApoLogies for the AmbiGuity.]
“Another staff member uses a variety of AI image creators to illustrate the stories, ensuring that the characters are relatable to Malian kids”
Once I observed online desperate struggle of a certain lady who wanted AI to create an illustration of an Ancient Egyptian dwarf without a beard…
Really? Ancient Egyptian dwarfs are canonically bearded? I had no idea.
If others had not been foolish, we should be so.
Some fifty years after I first read this, I think I have just this minute realised what Blake actually meant by it. Just goes to show …
(I have always been misremembering it as were not foolish, instead of had not been.)
Bill was quite right. Even if he was a bit eccentric at times.
Even though he was probably not an indigenous Egyptian deity, available images of ancient Egyptian dwarfs are probably totally dominated by Bes, who was overwhelmingly shown bearded.
Worthy of the Sino-Platonic Papers!
I think that’s unfair. (Heavily critical as I have been of some SPPs.) It’s Edo Nyland all over again. (All world languages derive from Basque, which was fashioned out of whole cloth by monks somewhere in the Sahara.) Heck there’s even a Hebrew diaspora all round the Sahara, ending up in Japan — because Eidelberg happened to visit.
” a certain lady who wanted AI to create an illustration of an Ancient Egyptian dwarf without a beard…”
This is what happens if you leave the keyword “eunuch” out.
And let us not forget A. Appathurai’s “Tamil and Sphere of Origin,” which demonstrates that all place names and ethnonyms are Dravidian.
This is on a level with telling English primary school children that their “native language” is Dutch.
Well, that honestly doesn’t seem that different from children in Bern being taught Hochdeutsch as their “native language”.
Or children in Guangzhou being told that Mandarin is their “native language”.
there’s even a Hebrew diaspora all round the Sahara
I have read in quite respectable authors that there may in fact have been Jews among the Arabic-speaking nomads who spread across the Sahara after the Muslim conquest of North Africa. (And of course there were certainly Jews in North Africa itself. Still are, just.)
Crank ideas about trans-Saharan Jews in an earlier time were once relatively mainstream, too. The Fulɓe were the usual preferred group; occasionally Soninke, in line with the then-usual idea that black Africans obviously could not have created their own civilisations. Bambara is a new one to me, but it fits the general racist pattern: if you’re a successful empire-builder, obviously you must have magical white genes somewhere. (Or if you can’t do proper Persil white, you can at least be “Hamitic.”)
One of the references in the Eidelberg WP article is to
http://www.kulanu.org/wp-content/uploads/magazines/2007-fall.pdf
which contains an on-the-whole-quite-sensible review of one of E’s works, by someone who doesn’t know anything about linguistics, and thus can’t see the batshittitude of the linguistic claims, but is properly dubious on other grounds. The review is interesting in listing quite a number of groups across the world who have decided that they are descended from the Lost Tribes (including some Igbo, I gather: though I know an Igbo man who really is actually Jewish, via his mother.)
Makes a change from all the African groups who have decided that they are Ancient Egyptians, I suppose. Or Qurayshi. Decolonise your minds, people!
I suppose eunuchs could be regarded as “transgender”, but in most cases it wasn’t something they had chosen, but something imposed on them. Maybe not so different, however, from the barbarities practised by institutions like the Tavistock Clinic in London (now closed down in the wake of the Cass Review), where children were encouraged by the trans lobby to think they were in the wrong “gender”.
If you visit Quora (something I don’t recommend, because it can become addictive) you can find plenty of questions along the lines of “Why don’t white people admit that Cleopatra and other ancient Egyptians were black?”)
Anyone who can speak of the “trans lobby” with a straight face has evidently never spoken to a single trans person
I would encourage you to get your head out of the heap of Tory bullshit where ideas like that grow, and look at (for example) the average time for a first appointment at a GIC is
Why don’t white people admit that Cleopatra and other ancient Egyptians were black?
I think that this has largely got to Africa itself as a result of Black American influence.
Traditionally, the thing to claim was that your illustrious forebears were Arabs of one kind or another. (Some of them actually may have been. After all, pretty much every European is descended from Charlemagne.)
The really interesting question, of course, is the Hungarian ancestry of the Saharan populations.
The connection between Hungarian and Kanuri is obvious to anyone who approaches the question with a truly open mind. (One can only marvel at the anti-Hungarian bias so evident in the so-called “linguistic” and “Africanist” literature. What is its true source, I wonder?)
I had no trouble accessing the archived version yesterday. Today it doesn’t work. Given that a completely different website also isn’t accessible today (…though in a different way: it doesn’t time out, it says “not found” immediately), maybe some Amazon Web Services server is down.
The article isn’t quite clear, but I interpreted it as saying that first the stories are generated in ChatGPT, apparently in French or maybe English; then Google Translate puts them into Bambara; “Experts like Tognine then correct any mistakes.”
1. Encourage children to think they’re in the wrong “gender”
2. ???
3. Profit!
I have been in a demonstration that was actually targeted by a counterdemonstration of feminist and lesbian activists objecting to the Welsh Government’s supposed trans-friendliness.
It was actually not too bad. Being feminists, they were not physically threatening, and we walked over and had some quite amicable conversations with them: I doubt if we changed anybody’s mind, but we probably did manage to undemonise the “trans lobby” a bit. I was glad I had some eloquent female comrades beside me though. Bad look for a man to be lecturing genuinely concerned women about these things.
I don’t belong to any “trans lobby.” I do think that no decent person should be trying to make the lives of an already marginalised group worse by spreading misinformation about them.
The lesson of the Tavistock thing is that you can do harm with the best intentions. I can reel off many horrendous examples of this in the context of overseas aid. Such things are bad in themselves, but worse because they give the Daily Mail demographic the fuzzy warm feeling that trying to help the poor is intrinsically stupid. They are wrong. Draw your own parallel.
Geopolitical realignment sweeping the Sahel in recent years (not just Mali, but Burkina Faso, Niger, & Chad) is increasing the number of Russophones on the ground there. They may *seem* merely to be gunmen or “advisors,” but maybe some of them will turn out to be crackpot amateur linguists who will help uncover a new and different set of hitherto undiscovered influences and substrates in the local languages?
@jwb
Nah. But these brave pioneers are preparing the ground for a special military operation in thirty years time, with the purpose of freeing them from Nato-backed Sino-African oppression.
Central Africa started teaching Russian in schools in 2019, I hear.
maybe some of them will turn out to be crackpot amateur linguists who will help uncover a new and different set of hitherto undiscovered influences and substrates in the local languages?
Crackpot linguistic theories do seem to appeal particularly to authoritarians, who often despise bourgeois “linguistics”, with its craven timidity in the face of actual evidence and failure to take proper account of Culture, Spirit and Will.
Why don’t white people admit that Cleopatra and other ancient Egyptians were black?
I guess a couple of millennia from now, some people will claim that Botha and de Klerk must have been black because they ruled an African country.
The Twenty-Fifth Dynasty pharaohs actually were (what Americans call) Black.
(Culturally, they were more Egyptian than the Egyptians. In fact, that was kinda their thing.)
“Nah. But these brave pioneers are preparing the ground for a special military operation in thirty years time, with the purpose of freeing them from Nato-backed Sino-African oppression.”
Russian military presence there makes me (it bothers me that DE calls himself “an amateur” while being an actual linguist, but I guess the definition “amateur” includes me…) less willing to do that.
Instead I direct my humble efforts at languaghat.com.
If I may return to the original article: I suspect, paradoxically, that this AI-produced book creation initiative will over the long run strengthen the position of French (and/or possibly English, Russian and Mandarin, depending on which way geopolitical winds blow) in Mali. If these books are indeed hastily-produced AI translations I would question whether monolingual L1 Bambara speakers would find them relevant or indeed even legible: if knowledge of French is required to make sense of a growing amount of written material in Bambara (and of written material in whatever other Malian languages books might be mass-produced in via AI), this will merely highlight to Malian non-French speakers the importance of mastering this language, and more broadly strengthen the perception that Malian languages in general are worse than useless.
PLUS ÇA CHANGE, PLUS C’EST PAREIL: I (and Michael Farris) had pointed out on this thread that script reform in ex-Soviet Central Asian countries was liable to (paradoxically) strengthen the position of Russian and English in that part of the world:
https://languagehat.com/nobody-is-really-too-keen-on-a-cyrillic-nta/
I also suspect a more immediate consequence of the loss of influence of France in the Sahel will be that Québec will be getting a lot of foreign university students from that part of the world. Nothing new under the sun: Back in my undergraduate days (in the late twentieth century, i.e. in prehistoric times from the vantage point of present-day undergraduates) my university had a large number of foreign students from North African countries who were personally quite hostile to France as a country but who mastered French better than any other language, and who thus crossed the Atlantic to undertake University studies (And for potential students from Sahel countries which are now getting politically closer to Moscow than to Paris, I can see the sales pitch already: Come study at a Québec University! Russian-like weather, without you needing to learn Russian!)
25th dynasty
Indeed. But kings and queens from most Egyptian dynasties would make much better poster children for Africanness than Ptolemaic Cleopatra. It’s just that nobody made a movie about them in which they’re played by Liz Taylor.
@ David E yesterday (sorry to be checking in late):
Bill had a hard row to hoe, IMO we should cut him some slack.
@Hans:
The fatale monstrum* deserves her celebrity. They done her wrong.
(She was also the only Ptolemy who could actually speak Egyptian, I gather.)
* Actually, she comes off pretty well in Horace’s ode, I reckon. Positive female role model. It’s Mark Anthony who gets the real dissing. Not even name-checked, just part of the grex turpium virorum.
@Jack M:
I’m pretty pro-Bill, on the whole. (I’ve even read “The Emanation of the Giant Albion” …)
Why did you take Vengeance O ye Sons of the mighty Albion?
Well, I can see why people might prefer to wait for the movie.
Terry Gilliam would make a lot of “all the Regions of Beulah were moved as the tender Bowels were moved…”
The Mental Traveler
As I said, I think they’re AI creations; the translation is done first by Google Translate (so more or less AI) and then corrected by L1 speakers, so it all hinges on how good the correction process is.
I think they’re AI creations; the translation is done first by Google Translate (so more or less AI) and then corrected by L1 speakers
Absence of heart – as in public buildings –
Absence of mind – as in public speeches –
Absence of worth – as in goods intended for the public,
Are telltale signs that a chimaera has just dined
On someone else; of him, poor foolish fellow,
Not a scrap is left, not even his name.
Undescribable – being neither this nor that –
Uncountable – being any number –
Unreal – being anything but what they are,
And ugly customers for someone to encounter.
It is our fault entirely if we do:
They cannot touch us; it is we who will touch them.
Curious from wantonness – to see what they are like –
Cruel from fear – to put a stop to them –
Incredulous from conceit – to prove they cannot be –
We prod or kick or measure and are lost:
The stronger we are, the sooner all is over;
It is our strength with which they gobble us up.
If someone, being chaste, brave, humble,
Get by them safely, he is still in danger,
With pity remembering what they were,
Of turning back to help them. Don’t.
What they were once is what they would not be;
Not liking what they are not is what now they are.
No one can help them: walk on, keep on walking,
And do not let your goodness self-deceive you;
It is good that they are, but not that they are thus.
whose work is this? thanks!
Auden. The Chimeras.
ah so. thanks
Hi. I’m Michael Leventhal, an, apparently, reviled person in your community from the comments I have read on the Washington Post article about our project to use AI in creating children’s literature in Malian languages. I’m honestly surprised by how many strongly held and perfectly uninformed opinions are current in your community, as well as your eagnerness to savage what you only know from a reporter’s article. There is too much nonsense in these comments for me to begin to respond … and I have better things to do … but let me just respond to the very first comment that we are using AI rather than humans to write the books. Nothing could be more untrue. Humans are using AI to aid them in the creative process, something like research assistants, and to do certain tedious tasks that the AI excels at like, especially in the creation of pedagogical material that accompanies each book and illustration. Every one of our books has a human, Malian author and is designed to delight children first and foremost, and to provide “mirrors and windows” for Malian children, that is, a reflection of their culture and environment so that they can identify with the content, and engagement with ideas and a wider world that may stretch their minds. There are numerous other aspects of our project which … as soon as we get a paper accepted somewhere … will be published. There is nothing political in our project … not even anti-French … that was political spin by the journalists. We have 65% illiteracy in Mali and we are trying to do something about it. The books themselves are publicly available in the Bloom library https://bloomlibrary.org/RobotsMali with English summaries. Judge for yourself. A little kindness people !
Hey Michael, thanks for commenting, and I’m sorry you felt picked on, though I’m not sure why — your name was mentioned once, in passing, in the post, and nobody was talking about you. Your project sounds great, and I’m glad to hear that “Every one of our books has a human, Malian author,” but all we had to go on was the reporting, which was (as usual) superficial and misleading. I’m glad you provided further information.
@hat, Michael
The reporter may have based some of his article on the RobotsMali website, viz.,
—
En tirant parti de l’IA, nous créons des histoires captivantes, des livres et des supports d’apprentissage en Bamanankan qui sont parfaitement adaptés aux niveaux de compétence des lecteurs en herbe. Les histoires sont ancrées dans la culture malienne, ce qui les rend pertinentes et attrayantes.
—
It is hard for me not to read that as assigning a primary role in the creation of the stories to AI, although I agree it could (just) be read the way Michael describes the process.
https://robotsmali.org/en/Projets/ai-for-education/
So the reporter made a total dog’s breakfast of it! As a scientist, I’m used to science reporting containing one or two fundamental misunderstandings, but I’m not used to total chaos… shame on the Washington Post, then.
https://restofworld.org/2024/mali-ai-translate-local-language-education/
This article is linked from the RobotsMali site itself. Evidently it (like several more) is seriously misinformed. Perhaps the site administrator should be told, so the link can be removed?
okay, i’ll be the killjoy.
Humans are using AI to aid them in the creative process, something like research assistants, and to do certain tedious tasks that the AI excels at like, especially in the creation of pedagogical material that accompanies each book and illustration.
so the plagiarism is more direct, and more traditional, being placed earlier and later in the process, especially in the “pedagogical material” (precisely where, in a language-education/language-support project, one might hope for things to actually be composed by an actual speaker – and trained teacher – of the language that’s intended to be supported).
that doesn’t persuade me (i don’t have the bambara to judge the results) to have any more faith than i did before that what’s being produced benefits the actual spoken language that’s involved, given the garbage that is presented as “english” through processes that are described identically. though it’s also worth noting that mr. leventhal’s account is the bog-standard (i suspect his automated “research assistants” might render that as a fen-, swamp-, or marsh-standard) fallback position of every Large Plagiarism Engine huckster when challenged on their claims, so it’s perhaps no surprise to find it reproduced anywhere along the ponzi chain where someone’s still willing to use the “AI” label about this software.
(In reply to Leventhal:) I am still in the dark. What are the incidental tedious tasks which are left for AI to do? Why does the project call itself “RobotsMali” if the robots/AI are marginal?
Why does the project call itself “RobotsMali” if the robots/AI are marginal?
They’re talking about opinion profit margins. “Robotics”, “AI” and “organic” are PR points. Small beer with big foam.
The impression one gains from the website is that their primary purpose is to collaborate with the government of Mali on teaching robotics and “AI” to Malian students; the basic-literacy stuff is not prominent. There’s more about translating textbooks into Bambara using Google Translate.
This focus is presumably based on an assessment that this should be a priority in Malian education. It’s supported by the Bill and Melinda Gates foundation, so we can be sure that this is in no shape or form a foreign-initiated project reflecting first-world concepts of how to support development.
By Morphic Resonance, I happened to allude in passing to this kind of model of international aid (in a different context) in this very thread.
Sadly the headline about their “Dispositif de guidage éléctronique intelligent” for the blind leads to a dead link. It does occur to me that it may not be a very sustainable (or scalable) model of how to help blind Malians …
The way I understood the article is that RobotsMali is doing what in the translation industry is called “machine translation post-editing” (MTPE). Which seems like a perfectly reasonable approach to me given the constraints.
That does indeed seem to be what’s going on with their translations of French textbooks into Bambara. (One can presumably be confident that the textbooks were not produced by ChatGPC.)
The main purpose of early education in West Africa is to get the child able to read French or English (or Portuguese or whatever.) Competence in the ex-colonial languages is extremely highly correlated with access to basic schooling. Most West Africans are pretty damn good at learning languages given the opportunity to do so (the motivation is pretty much a given, in this context.)
It would be lovely (speaking as one who delights in the myriad beautiful languages of West Africa, and would dearly like to see them flourish and be accorded all possible respect) if the real block to education in a country like Mali was that schoolchildren were held back by the lack of textbooks in Bambara. This is grotesquely far from the real situation. They are held back by lack of access to basic education. Every government franc spent on this kind of thing is diverted from better uses in training, paying and properly motivating schoolteachers, and providing a safe and healthy environment for schoolchildren.
Sorry, I was thinking about a different project when I wrote that last comment.
For the children’s books at least, RobotsMali seems to be using generative AI to create a rough draft, and then the human editor corrects and polishes that rough draft. Depending on the quality of the rough draft, this could save a lot of time.
There are of course groups like Boko Haram and its various allies or fellow-travelers who are, to put it euphemistically, vigorously critical of the sort of approach to basic education in West Africa that presupposes and rewards early fluency in a European-origin language. Although perhaps they are just trying to change the incentive structure toward incentivizing and rewarding early reading knowledge of Arabic rather than promoting local vernaculars.
Boko Haram might indeed be in the market for large-scale automated translation of English textbooks into (Ajami) Hausa. (Arabic might raise questions of genre appropriateness.)
One does rather get the impression that Boko Haram object to the content of English textbooks as well as the linguistic form, but then the current rulers of Mali seem happy to believe that the language used is more important that any actual distinctive cultural content (or complete lack thereof.) Perhaps ChatGPC (with appropriate guardrails, naturally) could generate acceptable materials?
The French have been ainfected.
#
Rencontrez notre équipe de journalistes spécialisés ainsi que des acteurs influents, des professionnels de la politique et des affaires publiques concernés par ces enjeux, à l’occasion de notre événement de lancement de notre couverture Energie & Climat France. C’est événement est sur invitation uniquement.
#
Energie & Climat France | Evénement de lancement
Depending on the quality of the rough draft, this could save a lot of time.
well, that’s the line the hucksters and shills are pushing, for sure. professional translators, who are the people with the most direct experience of what that looks like in practice, know that the opposite is true. they’ve been very clear that minimal time is saved, given the amount of correction involved to turn software-generated “material” into something actually readable in the target language and reflecting the source text. the use of Large Plagiarism Model software is very deliberately embraced and promoted by translation agencies and publishers as a means of cutting translators’ pay, or eliminating it entirely, not as any form of time- or labor-saving.
this kind of initiative, focused on applying that model directly to writing, has predictable effects. first among them is reducing (or eliminating) the possibility of making a living (or any meaningful amount of income) as a writer in bambara – which in turn damages the prospects for the cultivation and expansion of bambara literature, and thus (if we believe that reading and written language matter to language health) damages the language.
Exactly so.
I would think that the leaders of this project are by no means hypocrites and genuinely believe that they are helping the people of Mali by capitalising on all this business that the military junta has created by officially skunking French.
The medical director of the organisation I used to work for (who, like the organisation, had thought long and hard about how to provide aid which was both genuinely effective and could possibly have beneficial results in the longer term) used to visit, as part of his duties, many projects all over the world which were looking for support.
An all-too-common result was (he said): “These are good people. They are working hard. And they are doing harm.”
I’ve seen this myself over and over again. The problem almost always originates in rich-world individuals with a genuine desire to help, thinking up “solutions” to poor-world problems themselves in line with their own existing skill sets (“what do I as an individual have to offer?”), and then looking for someone in the poor country to collaborate with. (You can always find someone if you try. Who is going to turn down a rich source of funding from abroad?)
Sometimes the locals manage to divert the project into more useful channels. Usually they do not, and the power dynamic works strongly against them if they try. The project usually eventually runs into the sand because it is unconnected with the real concerns of the local people and their knowledge of the kind of assistance they need to address them. It collapses, or peters out, and the foreign personnel – along with the foreign funding – are pulled out. If you are lucky, at the end of the day, it will just have achieved nothing, but all too often it results in lasting damage from the undermining of local efforts to deal with the problem by your much-better-funded initiative. When you leave (and you will leave in the end, whatever you think) it is they that will have to pick up the pieces.
You can’t help someone properly if you don’t listen to them properly.
[That is harder than you tend to think, too. It was quite a while before I realised that in Ghana asking people who see you as the boss simply to express their own opinions doesn’t work if you’ve already given them some idea what you think yourself. Even appearing to disagree with the boss to his face is very bad manners. On the other hand, creative interpretation of what he meant, in line with what he ought to have said if he had really understood the situation properly, is fine. (Initially, I myself misinterpreted such sensible expedients for saving me from the consequences of my own ignorance.)]
C’est événement est sur invitation uniquement.
Is that something a French human might do, like English writers mixing up there and their?
I would think context-aware AI would in fact shy away from “C’est événement”.
Looks to me like automated mistranscription of spoken input.
Yeah, that occurred to me later as more plausible. Speech-to-text software even. In speech the written final “t” in cet is de-zeroed before a vowel, and the software restored the zero with c’est. if you believe in zeroes, anything is possible.
An AP (Automated Plausibility) assistant should be able to correct that text – or at least offer equally plausible alternatives, none of which is correct or colloquial.
in Ghana asking people who see you as the boss simply to express their own opinions doesn’t work if you’ve already given them some idea what you think yourself. Even appearing to disagree with the boss to his face is very bad manners. On the other hand, creative interpretation of what he meant, in line with what he ought to have said if he had really understood the situation properly, is fine.
Well, I have to disagree with that conclusion. About 20 years ago I worked at a Deutsche Bank project with many South Indian programmers. Many others – Germans and myself – were taken aback at the fact that these programmers couldn’t bring themselves to point out code errors or suggest improvements. It was hard to elicit possible reasons for this, as I found by talking with one or two of them.
Perhaps they couldn’t grasp the idea of error as something not merely “subjective”. They rarely volunteered an opinion, and preferred to take instructions. Very pflegeintensiv. SaaS (Submissiveness as a Service).
I have observed that in S-M activities the M side can be in control of the situation. For my autobiography I have drafted a chapter with the provisional title The Reluctant Sadist.
Whether my observations regarding Ghanaian mores have any bearing on South Indian culture I cannot say.
Bear in mind, too, that my Ghanaian subordinates actually did understand the general situation better than I did in many highly relevant respects. I suspect that the situation with your underprogrammers was not quite parallel.
Sadism is not all it’s cracked up to be. That James woman has much to answer for.
Not a Sahel-languages story but maybe this is a less-off-topic thread than other recent ones to stick this interesting news item into: The Newberry Library in Chicago (barely a block away from where I lived many decades ago) has received a substantial federal grant to digitize a portion of the Edward E. Ayer North and Middle American Linguistics Collection, viz “more than 2,000 rare books and manuscripts documenting [approximately 300 different] languages spoken or formerly spoken by the Indigenous peoples of North and Central America.” https://www.finebooksmagazine.com/fine-books-news/major-grant-funds-digitization-newberrys-indigenous-peoples-north-and-central
The illustration they give is of a two-page spread from Rodolphe Petter’s 1915 English-Cheyenne Dictionary, with the additional feature of the printed or perhaps typescript text being very heavily annotated in handwriting, possibly in multiple hands. Which may pose an additional technical challenge for digitization assuming the substance of those annotations is potentially of equal interest.