![[syndicated profile]](https://www.dreamwidth.org/img/silk/identity/feed.png){kind=small}
file770_feed - Pixel Scroll 8/23/25 He’s A Pixel Scroll Wizard, There Has To Be A Twitaljazeera_feed - Russia-Ukraine war: List of key events, day 1,277aljazeera_feed - North Korea’s Kim oversees test-firing of new air defence missiles: Reportatrios_feed - America's Worst Waylawyersgunsmoneyblog_feed - The Cannon canonscrooksandliars_feed - Florida Decides To Screw Over Kids With Disabilities Even Morecrooksandliars_feed - Imagine Being Forced To Sell Part Of Your Company To Trumpfile770_feed - Barkley – More Vignettes from Seattle Worldcon 2025aljazeera_feed - Nigeria says it killed 35 fighters in air strikes near border with Cameroonmikethemadbiologist_feed - Links 8/23/25crooksandliars_feed - Ukraine's New Pink Flamingo Missiles Strike Fear In Russiaatrios_feed - Afternooncamestrosfelapton_feed - AI, Marx, Labour Theory of Value, Vox Day etclawyersgunsmoneyblog_feed - If Democrats are going to win they need to stop saying things Republicans say they saycrooksandliars_feed - Orlandoans Bring Back The Pulse Rainbow Crosswalk, Defy MAGAalpennia_feed - Writing Projectslanguagehat_feed - More on AI Translation.aljazeera_feed - Tens of thousands of Palestinian children starving in Gaza tent campsaljazeera_feed - Oviedo vs Real Madrid: La Liga – teams, start, lineups![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
crown_of_antlers - Another Toad Cursetimeasmymeasure
As part of a good discussion of the searches and seizures chez John Bolton, Asha Rangappa reminds us of this particular episode in the death of the rule of law in America [gift link]:
In 2020, Mr. Trump’s Justice Department opened an investigation into Mr. Bolton to find out whether he unlawfully disclosed classified information in his book, “The Room Where It Happened.” During the years Mr. Trump was out of office, Kash Patel, now head of the F.B.I., also included Mr. Bolton on a list of 60 so-called deep-state names in the appendix of his book, “Government Gangsters.”
Soon after taking office this January, Mr. Trump terminated Mr. Bolton’s security detail, despite the fact that Iran had made threats on Mr. Bolton’s life and an Iranian was charged in 2022 in a plot to assassinate him. All of this has left the public with reasonable concerns that the justice system is being weaponized to seek personal retribution on behalf of the president.
At the same time, Mr. Trump has also attempted to delegitimize other, substantially similar investigations — when they were directed at him. He and other Republicans attacked the search when the F.B.I. executed a warrant for highly classified documents at his Florida home and private club, Mar-a-Lago, in 2022, declaring it a “witch hunt” and much more. The court released the F.B.I.’s affidavit, and the public was able to read for itself the substantial evidence underlying its probable cause showing. But that case withered on the vine following a series of irregular procedural and evidentiary rulings by Judge Aileen Cannon, a Trump appointee, further muddying the public’s understanding of the underlying national security threat posed by the facts of that case.
The classified information case against Trump was almost certainly the strongest, it was killed by the actions of a hack Trump judge so lawless her orders were mockingly vacated by panels of Republican circuit judges, and the mainstream political press more or less agreed to never mention it again, something I doubt will change very much even as Trump pretends to care about people illegally possessing classified information again. In fairness, it’s not as if they’ve ever been able to demonstrate sustained interest in an issue involving compliance with information security protocols.
The post The Cannon canons appeared first on Lawyers, Guns & Money.
As usual, with Republicans, cruelty is the point of whatever they do. The latest case in point comes from the swamp lands of Florida, where the Board of Education voted to make life impossible for some of their students who have the greatest needs:
The Florida Board of Education voted to sunset certificates of completion for students with disabilities who attend K-12 schools in the state. These certificates were awarded to students who couldn’t complete the coursework needed for a diploma.
Under a new Florida law, HB 1105, and the board’s vote, students with disabilities will no longer be able to get a certificate of completion at the end of their school career, starting this year. Students with severe disabilities who can’t earn a standard high school diploma, will leave school without any formal recognition upon graduation.
Without those certificates, some of these children are not going to have any chance of success for life. They won't be eligible for jobs, or even vocational or developmental training.
The Florida Board of Educations is just to ignore these kids while they pay attention only to the children they feel deserve it by not having mild disabilities or no disabilities.
A mother of a young man who had benefitted from the certificates of completion asked and then answered her own question about why Florida would do such an inhumane thing:
The US government has taken an $8.9 billion, 9.9% stake in Intel, buying 433.3 million shares in the chipmaker at a price of $20.47 per share. Via Yahoo News:
The government’s investment in Intel will be passive ownership, with no Board representation or other governance or information rights, Intel said in a statement. Intel said the government also agreed to vote with the company's Board of Directors.
Sure, except when Trump picks up the phone and tells you what to do!
Intel said the government's equity stake will be funded by the remaining $5.7 billion in grants previously awarded but not yet paid to Intel under the US CHIPS and Science Act and $3.2 billion awarded to the company as part of the Secure Enclave program.
Intel will continue to deliver on its Secure Enclave obligations and reaffirmed its commitment to delivering trusted and secure semiconductors to the US Department of Defense. The $8.9 billion investment is in addition to the $2.2 billion in CHIPS grants Intel has received to date, making for a total investment of $11.1 billion.
Links for you. Science:
Scientists Detect Unprecedented Energy ‘Tidal Wave’ from the Sun
Scientists fight back against Energy Department’s ‘antiscientific’ and ‘deceptive’ climate report
At 17, Hannah Cairo Solved a Major Math Mystery
Inside Jeffrey Epstein’s Forgotten AI Summit
A novel mobile genetic element with virus-like characteristics is widespread in the world’s oceans
Large Great White Shark spotted near Santa Monica
Other:
Former ‘Alligator Alcatraz’ worker describes ‘inhumane’ conditions inside
This conversation is being recorded — and so is everything else you do in San Francisco
This article on on NYC landlords unable to pay their expenses gives a foretaste of what I suspect the pushback to a rent freeze is going to look like (excellent)
The Death of the Fourth American Republic
A G.O.P. Congressman Faced Hometown Voters. It Wasn’t Pretty.
Tweens are about to feel the pinch of Trump’s tariffs
Post Fact Checker Glenn Kessler Exits in Disgust Alongside About 75 Colleagues in Latest Buyout
Where are all the Democratic donations?
Brown University is ‘functionally inaccessible’ to transgender students after Trump settlement
Trump may be off the hook for his 2020 election plot, but his allies aren’t
Brown’s Trump surrender shatters a dream
How rethinking stairwells could boost housing in Montgomery County
In draft congressional map, Texas Republicans bet big that gains with Latino voters will persist
On Redistricting
DC could be a garden city, with healing effects
The rewards of ruin: Societal downfalls loom large in history and popular culture but, for the 99 per cent, collapse often had its upsides
Christian Nationalism’s Plot on Civil Society: The Seven Mountains Mandate
Fighting Wildfires Is Hellish Work. It’s Even Worse Under Trump. As the Trump administration’s climate denial collides with its quest to incapacitate the federal workforce, wildfire fighters are taking a hit.
These Democrats Think the Party Needs AI to Win Elections (it’s bullshit all the way down…)
Why Your Team Sucks 2025: Dallas Cowboys
The eternal vacuity of reactionary centrism
Congressional Watchdog Publishes Aptly Timed Reminder That WH Budget-Slashing Scheme is Illegal
MAGA Diehards Say Loomer is on Big Pharma’s Payroll
A Big, Upcoming Fight Could See Dems Demand Congress Take Back Its Power
Women shouldn’t hold office, says GOP woman now running for office
Ignoring scientists’ pleas, RFK Jr. slashes investments in mRNA vaccines
Utah’s Senate president prompted law change that helped a teen charged with child rape (PedoCon theory at work)
RFK Jr. Cancels $500 Million In Funding For Vaccine Development
A Look Inside Jeffrey Epstein’s Manhattan Lair
"A factory error left early FP-5 cruise missiles pink, leading Fire Point to nickname it “Flamingo” — a name that stuck." The company's director, Iryna Terekh, an architect before the war, liked the color though, so she kept it for the nose. "Woman's touch in missiles," said Terekh.
When she was hired, the goal was to produce 30 FP-1 drones a month. Fire Point now claims to be producing up to 100 per day. And somehow managing to do all that for just $55,000 a piece. It has many wondering how Ukraine can do all this for practically nothing, and the FP-5 at a fraction of the price of a similarly specced US-made Tomahawk missiles, which cost over $2 million each
Source: Politico
KYIV — While Ukraine's allies twist themselves into knots over how to provide it with security guarantees, Kyiv is moving ahead with its own way of keeping Russia at bay — a new cruise missile able to hit all of European Russia carrying a massive warhead weighing more than a ton.
That's the Flamingo FP-5 missile, developed by Ukraine's Fire Point defense company — a fast-growing producer of the combat drones that have become a key weapon in the war against Russia.
There appears to be evidence that over use of LLM chat bots rots your mind but don’t worry, Vox Day was always like that.
First though, in the spirit of always burying the lede, here is where I think we are on the issue of LLMs as a political spectrum.
The political dynamics of LLMs is interesting. On the left, a consensus arose that giving even an inch on use of LLMs is a betrayal of writers and artists. The left backlash against LLMs et al can seem extreme but given the huge corporate pressure from technology companies for people to adopt these products, it is a proportionate response. Companies like Microsoft see products like ChatGPT as a way of vastly increasing use of their cloud based services and have invested huge amounts of money in the expectation of “AI” being integrated into everything.
The left opposition is pretty easy to make sense of. Big corporations want to make huge amounts of money out of a scheme that is based off charging rent and smaller corporations hope to be able to sack a lot people and replace them with AI. The social and environmental cost of that plan will be high and it really is straightforward how the left (in general) has ended up being very sceptical about AI, even people like myself who are intrigued by how the shiny computer thing works. There are a pile of other issues with the current AI boom/bubble but the why of specifically left opposition to LLMs etc is very simple.
Meanwhile, writers and artists have a related issue. Their work is both being replaced by these models AND their work was effectively stolen to make the models. For writers and artists on the left that is a quadruple whammy of reasons to be opposed to the ChatGPT era. However, for writers and artists not on the left there is still clear commonalities between the more overtly anti-capitalist aspects of AI-opposition and the pro human creativity aspect of AI-opposition.
For people less consciously ideologically driven who are not currently having their livelihoods threatened by LLMs, ChatGPT is often perceived as fun and potentially useful tool. I’ll call this the naive centre. There’s not necessarily a great understanding of the limitations of these models, nor the social cost of their use, nor the current weird economics of them (basically, this is an industry speculating on future profits rather than one currently making a profit). Meanwhile, governments are being largely pro-AI on the grounds that what is good for business must be good for society.
But what about the Right? Modern right-wing politics can swing wildly from pro-corporate positions to very anti-corporate positions. The influence of people like Elon Musk on the US right has meant the pro-corporate side of the right has tended to win out on the issue. There is a simmering this-is-new-and-weird-and-hence-demonic view of AI within the right but it isn’t dominant and has been ameliorated by the arrival of perceived “non-woke” chat bots like Grok.
But what about right wing writers and artists? They are in the same boat as all writers and artists. Their work was effectively stolen to feed these models and is then being used to create content intended to replace the work they do. What they do not have is any kind of community solidarity on the issue, they don’t even have solidarity between right wing artists and right wing writers. You will see writers in right wing spaces bemoaning the writing of GPT as AI-slop who are, nonetheless, happy to use Midjourney to make their own book covers.
Even so, within right wing creative spaces (and I am using the term ‘creative’ generously) there is increasing opposition to AI. If you are somebody who was hoping to supplement your retirement income by selling a bunch of books on Kindle Unlimited, then the market being flooded with machine generated books of comparable quality is a direct threat. As many people have been discussing in the comments, ChatGPT is not a viable way of writing a good short story but it may be good enough to churn out low-quality by-the-numbers military science fiction. Worse, even if ChatGPT isn’t good enough to do that, there are enough grifters to flood the market with crap that will drive readers away. LLM’s are a bigger existential threat to the business model of writing on the right than they are on the left.
So, rough summary:
Vox “not a Nazi” Day is somebody who has both IT credentials and writing/publishing credentials. They are not particularly GOOD credentials but he has genuinely worked in computer game development, book publishing and has written fiction and non-fiction books. Arguably, the quality of his writing is so bad that maybe using LLMs would have been an active improvement. Notably, he wrote or maybe co-wrote (or at least published) a science fiction ‘novel’ in which a future society becomes so dependent on AI that when the algorithms start becoming self-corrupting it leads to a societal collapse. That book is so bad that while I was reading it, I had to read a Wikipedia article on packing peanuts just to stop my brain from atrophying.
Day’s relevance on the right has been declining sharply for years but he is now refashioning himself as a champion of AI within what we used to call the Alt-Right. He has a dedicated Substack on AI and his own blog has been pushing AI writing, AI images and AI music.
He has also recently posted on his blog a framing of AI in ideological terms. Entitled “Opposing AI is Marxian” [Archive Link] the post claims that opposition to AI is inherently Marxist. Here is the introduction:
“Since it’s obviously too difficult for the average individual who denigrates the use of AI and opposes its use on the grounds of insufficient human involvement to understand their own position well enough to recognize its obvious intellectual roots, I asked Claude to dumb down my observations enough to permit their little midwit minds to grasp it.”
https://web.archive.org/web/20250823190058/https://voxday.net/2025/08/22/opposing-ai-is-marxian/
“Claude” is a large language model made by the company Anthropic. You can read about it here on Wikipedia https://en.wikipedia.org/wiki/Claude_(language_model) 1
So, yes, TECHNICALLY Day didn’t write the post an LLM did but he posted it and claims it correctly summarises his argument.
“The Hidden Marxism Behind “AI Slop” Complaints
When critics dismiss AI-generated art as “soulless pablum” or “AI slop,” they’re often unknowingly channeling a 19th-century economic theory that most economists abandoned long ago. Their argument, stripped to its core, reflects the labor theory of value that Karl Marx popularized—the idea that something’s worth comes from the human work put into it. This perspective, while emotionally appealing, fundamentally misunderstands how we actually value art and creativity.”
Oh boy.
So, yes, there is an argument going around that claims that AI art or writing is intrinsically bad because it lacks either soul or a human connection etc. I think it is a weak argument but there are ways you can make it stronger and like a lot of points made on social media, it is often a quick way of describing a more complex idea e.g. that art of all kinds is part of a social dialogue. However, Day (or Claude or the Claude-Day hybrid cyborg) is looking specifically at what might be a straw man version of the argument i.e. AI generated stuff lacks aesthetic value because it lacks human soul.
Let me note something straight away about that straw man argument. It is fundamentally NOT Marxist in nature. I don’t mean “no Marxist has ever said anything like that” or “this is not an argument people on the left use” because people say all sorts of stuff. No, what I mean is that the argument as presented above is fundamentally NOT a materialist one. Instead it relies on spooky essences (hence ‘soul’). Yes, there are ways of framing it in a social sense but that’s not the what Day’s post is looking at.
Secondly, the argument is:
This post is already too long, so I am not going to attempt a full treatment of the labour theory of value. Short version is that classical economist such as Adam Smith speculated that goods gain monetary value based on the effort people put into making them. That makes intuitive sense as at least one factor but there are lots of counter-examples and factors such as scarcity play a role. Economists aren’t idiots and so a naive labour theory of value (i.e. goods are worth money purely based on the effort needed to make them) isn’t really a thing. Karl Marx also was not an idiot and didn’t believe that either and more to the point didn’t even use the term “labour theory of value”. He did argue that capitalism exploits the value created by labour but note that is a long way from saying that the only source of value is from labour (indeed it overtly implies the opposite).
So:
What Day is presenting, if I were to be generous, is an argument from analogy or, if I was to be less generous, just getting really confused because the word “value” has more than one meaning.
There is a tradition of left wing thought that celebrates the aesthetic/cultural value from the labour that goes into making things. However, that tradition is (as explained above) not the Labour Theory of Value but also it is not an exclusively left wing tradition. The idea that carefully made physical goods have an intrinsic quality to them is an idea that you can find within small-c conservative thinking and even more right wing traditionalist thinking.
For example, I can think of a far-right ideologue2 who also runs a business crafting expensive leather bound editions of classic books. If you visit his websites, when he talks about these books he emphasises not just the quality of the material but also the care that went into choosing the materials and the methods used to create the books. Now, maybe the books are actually dog shit, I wouldn’t know but even this is all just marketing spin, this is an anti-Marxist selling stuff to other anti-Marxists using appeals to what we might call “the labour theory of aesthetic soul”.
So here is the question: is the blog post written by the Claude LLM so bad becuase LLMs are automated bullshit creators and no human effort went into the construction of the argument? Surprisingly, no. I find Claude not guilty of this charge! No, the argument Vox Day posted is so bad because he has always been like that. LLMs didn’t rot his brain because they didn’t need to.
︎︎
The Third Way, a think tank great at getting media write-ups but not evidently providing any other “value,” got predictable positive coverage for circulating a list of “woke” terms that Democratic politicians are purportedly using to alienate voters. My reaction, like a lot of other people’s, was to ask which Democrats running for national office were using any of these terms:
i consume an extraordinary amount of political content and i have never seen someone in elected office use the term "birthing people"
— jamelle (@jamellebouie.net) Aug 22, 2025 at 8:53 AM
I was given a single example, from a short-term house backbencher who lost in a primary. But who knows, maybe Dems running for office are using these terms more than I’ve noticed?
Helpfully, Lindsey Cormack has stepped in with a systematic study of the question:
I don’t know anyone at Third Way and had never encountered that group or website before, so none of this is “personal”, but when someone asks me to look at congressional word usage – I’m pretty much always up for it.
I agree that many of the words listed have more negatives than positives in political arenas. But the idea that Democrats have been the ones driving these terms into the public sphere is worth checking against the evidence. Terms like “woke,” “critical race theory,” or “diversity, equity, and inclusion” didn’t come out of Congress, they migrated from leftist activist or academic spaces…and then were repeated endlessly by Republicans more-so than Democrats, but in a way to complain that Democrats had been using them too much. In fact, this is the topic of a book I’m working on (and if that’s interesting to you, I’d be happy to hear your thoughts on the matter).
Here’s how it breaks down when we check the DCinbox archive of 208,000+ official congressional e-newsletters from 2010 to today to see who uses the words and terms outlined in the Third Way memo.
You can click through for the results, but a lot of the Third Ways boogeymen terms have no appeared in a single congressional email since 2010, some are used not very often and similarly by members of both parties, some are actually used by Democrats more but generally rarely, and some are used almost exclusively by Republicans. The latter is perhaps the most telling category — terms like “birthing person” from the Third Party list are terms used only by Republican politicians to falsely accuse Democrats of using them. The only member of Congress to use the term “microaggression” or “cultural appropriation” is Matt Gaetz.
Cormack’s conclusion:
Looking at actual usage, the Third Way memo reads less like an audit of Democrats’ language and more like a list of terms Republicans tell us Democrats are saying. The data show that many of these phrases barely exist in constituent communications, and when they do, Republicans are often the ones writing them either to lampoon Democrats or to spotlight them as proof of “wokeness.” But again, these are not campaign emails, and I’m far out of campaign world for the most part.
But in doing this version of a check and in my understanding of how American politics can move forward in a more functional way, I agree we need to get away from what Third Way calls “the eggshell dance of political correctness.” People and politicians should be willing to adapt words when they don’t land and should be open to trying out new terms that capture novel experiences/problems that we need to deal with.
But as long as Republicans can keep defining Democrats by terms Democrats themselves rarely use, and everyone comes to believe this through repetition is a much bigger challenge for the impressions of the Democratic Party than any lefty words they might on occasion.
I’m honestly not sure what to do about the general media consensus that Democratic politicians are responsible for everything ever said at an obscure academic conference, but Republicans are not responsible for the rhetoric of Republican presidents. But one thing that could help a little is for organizations that are supposed to be helping the Democratic Party to stop circulating lies about what Democratic politicians are saying to help the media portray Democrats as out-of-touch elitists.
The post If Democrats are going to win they need to stop saying things Republicans say they say appeared first on Lawyers, Guns & Money.
As C&L reported last month, former Fox News host and current Transportation Secretary Sean Duffy urged governors to get rid of rainbow crosswalks, even though he had to admit that traffic fatalities had declined in 2024.
Duffy pretended he did so in the name of safety but, like everything else the Trump administration does, it’s always MAGA politics first and American safety wherever it’s most expedient afterwards. Duffy barely tried to hide that in his social media post when he wrote, “Political banners have no place on public roads ... Taxpayers expect their dollars to fund safe streets, not rainbow crosswalks." Taxpayers expect their dollars to fund vaccines, medical research, and national parks, too, and to be told why our president is in the FBI files of sex-trafficking pedophile Jeffrey Epstein, too. But I digress.
It was no surprise that Gov. Ron “Don’t Say Gay” DeSantis jumped at the opportunity to erase a memorial to the 49 people killed in a 2016 mass shooting at the gay Pulse nightclub, in Orlando.
But Orlandoans responded by refusing to allow that piece of history and commemoration to be erased. They showed up in large numbers, many caped in rainbow flags, to restore the rainbow crosswalk with chalk.
I promised that one reason for taking a short vacation from LHMP blogging was to get traction on some other writing projects. To start with something bite-sized (because it's always nice to get that boost from completing something) I went back to a short story about a harpy who's moonlighting as a muse. Following my plan to try writing long-hand to avoid distractions, I got three pages written. And then when I went to transcribe them, I discovered that I had a lot more already written than I remembered. (This is not an uncommon experience for me.) I'm approaching it with a different tone and voice this time, but it meant I could pull a bunch of existing text and plot-points. However that also meant that I needed to work on it in Scrivener rather than on paper. Hey, whatever works. Ironic that I started with a story all about writer's block and inspiration?
My other non-LHMP project at the moment is a deeply geeky history and analysis of the Best Related Work Hugo category. In part, because I wanted to try to answer the question "what exactly do fans think is a related work?" and in part because I like having something analytical to sharpen my brain on. The first stage of the project involved tracking down all the available nomination data (the long-lists, when available) and coding items for format, genre, subject, etc., as well as confirming correct (full) titles, author/creator names, and publishers. Currently I'm confirming author/creator gender for statistical purposes (which in a few cases means "gender at time of authorship or current gender?"). I wish I could do the same for cultural/ethnic background but that's less likely to be referenced in biographical material.
The other initial part of the Best Related project is tracking down the administrative history of the creation and revisions to the category, including discussions of intent. Plus doing the same for other Hugo categories that have overlapped in some way (e.g., where a new category was created that subsumed material previously falling under Best Related). At each step, I'm making notes of new rabbit holes that need exploring. The most annoying gap is that for the earliest stage of the category (1980-1997 when it was named "Best Non-Fiction Book") expanded long-list nomination data is only available (at the official Hugo website) for 4 years, and the number of long-list nominees is variable for that data. I have a suggested lead on a possible archive that might have Hugo-related ephemera, but I'll have to decide how deep I want to dig. Currently the available data corresponds nicely with the changes in category name (for the "Best Related Book" and "Best Related Work" eras, full long-list data is available) so I have sufficient detail to do some valid comparisons between eras.
The eventual result of this project will be published here as well as potentially in some location with more traffic, but it will take a while.
Victoria Livingstone writes about the ever-more-pressing issue of using machine-translated texts to save money:
I lived in Latin America for several years and I speak Spanish fluently, but I am not a native speaker. I proofread translations into English and my co-worker, who was a native speaker of Spanish, proofread Spanish. Together we were in charge of quality control for that language pairing.
We once received a machine-translated document that included the phrase, “HIP’s asthma program.” HIP was an acronym for “Health Insurance Plan,” but Google Translate (in a document sent by one of our clients) rendered the phrase as the colorful and absurd “asma de la cadera” (quite literally, “asthma of the hip.”) Machine translation has greatly advanced since then. I just put the same phrase into ChatGPT and even without the full context of the insurance plan brochure, the model returned “el programa de asma de HIP.”
What about more culturally charged phrases? My co-worker and I were once tasked with translating text into Spanish for a televised notice on water pollution. “Imagine water pollution as rubber duckies,” the ad began. It was accompanied by an image of thousands of swirling yellow ducks. My colleague pointed out that rubber duckie is a culturally charged term. It is iconic as a toy in U.S. culture. My co-worker was from Mexico City, and to her ear, using “patito de hule” (or something similarly literal) as a central analogy was bizarre. This was years ago, but I believe we translated the term more generally as “juguetes” (“toys”). Today I prompted ChatGPT to translate the phrase, and it returned “Imagina la contaminación del agua como patitos de hule” (a grammatically correct but uninspired rendering of “imagine water pollution as rubber duckies.”) The AI-generated translation, then, worked well with the asthma program but not with a culturally charged metaphor.
The UK-based company GlobeScribe.ai recently announced that for $100 a book, they will translate fiction through AI. The translations may be literal or superficially competent, but I suspect that style and nuance will be flattened in AI-generated translation. What would they do with rubber duckies?
In “Culture is Not Trivia: Sociocultural Theory for Cultural NLP,” researchers Naitian Zhoul, David Bamman, and Isaac L. Bleaman note that LLMs draw on datasets that a “are limited to static snapshots of cultural artifacts.” Culture is “not a collection of trivia,” they argue, but complex, dynamic, and challenging to define. In a recent article for The Observer, translator Daniel Hahn gives the example of translating the simple phrase “T’es fatigué, toi?” (“Are you tired?”) from French. While simple, the phrase contains information about the gender of one of the characters and, in the context of the literary text Hahn was translating, communicates a shift in the nature of the characters’ relationship (from more formal communication earlier in the text to the informal use of “toi”). Translators, as Hahn notes, interpret phrasing and make informed decisions as to which elements are most important to preserve. They draw on linguistic as well as cultural knowledge. […]
In the New Yorker article “Why AI Isn’t Going to Make Art,” Ted Chiang writes that “art requires making choices at every scale; the countless small-scale choices made during implementation are just as important to the final product as the few large-scale choices made during the conception. It is a mistake to equate ‘large-scale’ with ‘important’ when it comes to the choices made when creating art; the interrelationship between the large scale and the small scale is where the artistry lies.” LLMs, Chiang argues, may draw on a vast amount of data but the models are not making thousands of small-scale choices in the same way a painter or author does. The same can be said of translation.
Of course, it doesn’t matter how lousy the result is, just as it doesn’t matter how ugly the new buildings that surround us are — profit is all. History to the defeated/ May say Alas but cannot help or pardon. Thanks, Y!
