Meandering Muse

TV Appreciation Week 2024  : Day 7 : Free choice → Favourite show nobody watched (Galavant)

(Or it isn’t; as you rightly point out in the tags, it’s a capitalism problem. But I’ll expand on this point of “capitalism”. I need to rant. I need to scream.)

I’m a professional translator. I work in video games and software, with an occasional dash of literary translation. I’ve worked in translation proper, I’ve worked on editing other people’s work, I’ve led a couple of translator teams. I’ve worked the occasional miracle, working around some Really Dumb Choices the developers made.

(Spoiler alert: other languages have different syntax and grammar, if you give me a list of nouns to translate, and then give me the plural “s” to translate separately, this is not good. Even in English, woman -> womans is dumb.)

I am a fan of making things affordable and accessible. I am really happy that Google Translate and similar things can tell me the gist of what people are saying in conversations I only half care about. As the poster above says, it’s great! Not perfect, but ok!

Do you know what’s not great? Do you know what the OP in the original image means?

The client the original image is talking about isn’t you. It’s not some person on the internet trying to find out what someone said in a Post. The client they’re talking about is, essentially, the corporation: the translation agency, the publishing house, the IT giant.

You, the individual, do not have the power to demand how I do my job. If you come to me and say, “Sarshi, I want you to take this 300-word post, run it through Google Translate, and then charge me half of what you usually do for translating it”, I can take it or leave it.

But I get contacted by agencies - half of them want this. “We have a game, Sarshi! Just post-edit the results of a machine translation!” “We have support articles, Sarshi! We’re paying you a lot less to post-edit the results of machine translation!”

You say it’s ok to have 80% accuracy, and I feel you! Yes, sometimes it is! But companies are like “lol, this works”, too!

It’s happening over and over. And these aren’t… they’re not people, you know? They’re not Auntie May trying to figure out what the dough recipe she got from her niece in Indonesia says. They’re agencies, trying to increase their earnings by promising top quality to companies, then going, “gosh, we said we’d do it for cheap, how can we manage that?”

Or they can even be large companies themselves. Oh, you’ve spent a bajillion trillion dollars trying to create the CryptoNFTVirtualRealityAI hybrid that everybody knew wouldn’t work and now you panic because your earnings are lower than usual? Oh, and you want to “cut costs” by screwing over every contractor you have? Great. Just great.

This is going to screw you over - you, the individual. Not my client, not the translator’s client in general - the company’s client. The corporation is too big to really care about how you feel about their product - the employees individually might, but the company’s only metric is if you buy it or not. And the company makes decisions based on what brings the most money for the least cost.

So your hardware manuals might be crap and you might be in tears because you have no idea how to make your new appliance do the thing. You’ll go on YouTube and you’ll find a solution, and you’ll eventually figure it out. And maybe you’ll forget about the crap manual in time. So next time, they still won’t get a good translator, because they already have a cheaper solution that seems to work.

So your game looks like it was translated by a bunch of rats in a bunker and you can barely understand what anyone’s saying? Well, maybe they got a bottom-feeding agency overpromise that they totally have legit translators working for $1/hour. Pinky swear! Did you buy the game? You did. So… the system worked! They’ll hire the same agency again!

It’s like the clothing industry all over again. We could have better clothes, but it’s cheaper not to. They’re doing us a service by selling us shoes that won’t last a season, and T-shirts that will look like crap after washing them twice - they’re cheap, aren’t they? They’re affordable. Anyone can get clothes. (So you pay more in time are are more frustrated? Who’s counting!)

And meanwhile, it’s easy to forget things might be different. That we have the ability to create good things, pleasant things. That manuals can be easily readable, that games can sound great, that books can be awesome to read. It becomes harder to trust the market, harder to believe in quality, easier to say that this is normal, this is how things just are.

And if you speak English natively, well… You’re at a huge advantage. A lot of stuff is created by your people, for you. For countries like mine, that are small enough to import a lot, nearly everything is translated. I want you to imagine almost all movies subbed, every appliance made elsewhere (with menus needing translated and all), every app in a foreign language. And everybody who can cut costs will try to.

cause like. apart from the (potential) exception of writing code, there is not really anything useful that a Large Language Model can do.

it imitates patterns in the text that it takes in

that's fundamentally how it works

Yes, haha funny, the AI is bamboozled by spelling, but seriously: I asked Google the same question 5 different times just now, and not only did I get 5 incorrect answers, I got 5 completely different incorrect answers.

The fact that they're different every time demonstrates just how much of an LLM's output is up to random chance. This is how LLMs imitate human language: each stage of constructing the text has built-in randomness, just like how human language has variation and flexibility.

This is my problem with using these things in the workplace. Like, imagine you're using an LLM to fill out a report of an accident that happened, or to write dietary recommendations for a patient, or to give feedback on a paper a student wrote...really anything you can think of.

Even if the LLM-generated text doesn't say anything technically flat-out wrong, the words and details are still randomized instead of being chosen by you for an actual reason.

Zan Zan’s masterclass: A-YAO and his 3 smiling modes

可利用  kě lì yònɡ - lit. available, usable      (maybe, fawning)

可抹杀 kě mǒ shā - obliterating

The Gothic Pool of Landa, Burgos, Spain 

Last week, the Trump administration quietly released a sweeping new federal rule that would use funding threats to force institutions across the country to reject transgender people. The 400-page proposed regulation would codify the administration's anti-trans executive orders into binding federal policy, imposing a blanket prohibition on federal funds going toward "gender ideology"

The proposed rule, formally titled "Regulation for Federal Financial Assistance," rewrites the government-wide framework governing all federal grants across every agency. Among its most consequential provisions, it requires that before a federal grant recipient can receive money, the award must pass a "pre-issuance review" conducted by a political appointee—not a career expert or peer reviewer—to ensure it is "consistent with applicable law, Federal agency priorities, and the national interest." The regulation explicitly instructs these appointees to screen for "denial by the recipient of the sex binary in humans or the notion that sex is a chosen or mutable characteristic." [...] An institution that acknowledges transgender people exist—through its policies, its training, its healthcare, its bathroom access, its HR procedures, its name-change processes—could be deemed to "deny the sex binary" or to “support the notion that sex is mutable” and have its federal funding blocked.

This torches everything. And I do mean everything.

A very abbreviated list of its ramifications include (but are not limited to):

ending funding for ALL DEI related initiatives

allowing the government to terminate grants at any point for any reason

preventing researchers from publishing, going to conferences, and being part of academic societies

requiring that topics must support the president’s agenda.

What this means, and if anything I’m under selling it, is the death of science and research in America. It allows the government to restrict any topic they please at a whims notice, putting officials who have no background in the topic in charge of deciding funding continuity. It controls what gets researched and if/how researchers are allowed to share their discoveries. There are no books to burn if the government never allows them to be written. This is fascism plain and simple.

Public comments are not a vote, it is a chance for you to say "here is an issue with this law I think you need to address" and provide justification for legal challenges if it goes forward:

"Comments raise the bar that agencies have to meet when making a rule; “if an agency fails to adequately respond to significant, relevant comments in a final rule, members of the public may seek to challenge the rule in court on that basis and claim it could be struck down.ˮ"

This year’s is different. They are now federal REQUIREMENTS, which means there’s no flexibility. As was said previously, the 400 pages are not singularly devoted to being absolute shitheads to trans people. Theres a lot of stuff in there, some of which is the standard dry boring grants stuff, some of which is the horrible ideological warfare outlined above.

This document is issued by the OMB, the Office of Management and Budget, which is currently lead by fucking Russell Vought, the principal architect of Project 2025. This is how they’re going to implement all the horrible shit in there that wasn’t covered by Executive Order. Russell Vought is actively coming for my job, my marriage, and my kid, and most of my friends lost their jobs last year because of him. He is the fucking arch villain behind the heinous shit the current regime is doing.