i love you archival work. i love you alphabetizing. i love you sorting. i love you reshelving. i love you document restoration. i love you shelf reading. i love you inventorying. i love you analysis. i love you archival work.
alphabetizing. analysis. archival archival document i i i i i i i i i inventorying. love love love love love love love love love reading. reshelving. restoration. shelf sorting. work. work. you you you you you you you you you
Japanese is an incredibly fun and rewarding language (if you’ve ever wanted to learn it for ANY reason, most importantly including the “silly” reasons) but the fearmongering and capitalist intervention involved in the language learning process have given it a reputation as an “impossible task” for English speakers, leading to confusion and dkn learners and weird pessimist attitudes about the whole thing. In this thread I will explain how to effectively learn and retain Japanese. This is a tried, tested and true method; probably 99% of all people who try to learn Japanese give up, but everyone I’ve met who has tried and stuck with this has been at or above N3-N4 after 6 months or less including me
You can teach yourself Japanese for free if you have a little free time every day and a computer
1. Drill yourself on hiragana and katakana. These are the phonetic building blocks of Japanese, think of them as equivalent to english letters. This site is a good resource in general. Once you have a solid grasp on this, DO NOT LINGER HERE; move to step 2. You will master kana later.
2. Download Anki. This is a flashcard service. They have a paid app if you’re willing to invest for it, but if not, they have a mobile website (create an account and sync it with your computer).
This is the deck you’re going to download. Import it to Anki and do this every day. I have learned the hard way (twice) that skipping this is bad. If you become overwhelmed, you can change the number of new cards and reviews by clicking the cog next to the name of the deck!
3. Cure Dolly (Youtube, grammar) + transcript. She has kind of a posh accent, you might want to turn subtitles on. Watch a few videos when you feel like it but most importantly set up 4 and 5 as soon as possible
4. Yomitan (must have) is a browser extension that functions as a pop-up dictionary. you need to install dictionaries for it to work. here are some dictionaries you can use with yomitan and explanations of what they do
5. READ. DO NOT LET YOURSELF GET STUCK BEFORE THIS STEP. JUST READ!!!! Most people who fail to learn Japanese do so because they are afraid of not being ready to move on, which is counterproductive. Just read. When you were a child did you spend years on vocab and grammar before reading? No I bet you did not. Pick something to read and learn what you don’t already know by reading in Japanese.
Jiten.moe has a list of novels and visual novels that you can read on your computer sorted by difficulty. So does jpdb. There’s also this document. There’s also this document. Hey look this website is cool too
For visual novels: download LunaHook. It “hooks” to your VN and allows you to use Yomitan on words you don’t know. Turn off the translation feature, it does nothing to help you learn
For literary texts: ttsu e-reader supports epub and htmlz files.
You can also learn Japanese by watching anime, but it’s a little more convoluted and requires a lot more patience.
For manga, utilize Mangatan, but I don’t recommend this right out the gate because when you’re first learning sentence structure you’ll want something with complete sentences.
Set your computer up for mining vocab before you start reading. Once you finish your kaishi deck, you can drill your mining cards (I didn’t do mine until after finishing kaishi because it was too much).
Most importantly: reading is going to be hard at first. It is going to piss you off. You need to muscle through with this because this is where the bulk of your learning will happen. After a while you will just feel like reading because you love reading! Try not to pick something too hard for your first read, but if you’re interested in the story you might be able to muscle through something a little tougher.
Remember to consult yomitan and cure dolly where needed, that’s what it’s there for. As you can see I am quite normal about the Japanese language, so if you have any other questions or need help with anything else feel free to shoot me an ask and I will get back to you promptly. Japanese is not your enemy and it is not impossible. It is your friend
this is true, ive been trying to learn japanese on and off for over 10 years, and the past year is the only time ive been able to actually stick with it every day and actually interact with the language meaningfully! here's some more resources:
1. the moe way i cannot overstate how essential this site was for me to change how i approach learning japanese. it not only has a plethora of resources to get you to start learning but also HOW to learn and why people tend to fail. i recommend reading their "japanese guide" and "30 day Japanese" to get you started. they also have a discord server with lots of extra resources. there's a lot of people there fluent or near fluent and native speakers too, so they can have some great insights for you
2. this is gonna age me severely but i recommend for learning the kana trying out this book too: Remembering the Kana: A Guide to Reading and Writing the Japanese Syllabaries in 3 Hours Each. it's only like 8 bucks for the digital version but it's so old the pdf has been circling around forever if you can't pay. they are not kidding, over 10 years ago when i first used this i really did memorize it in a few hours, and it took like 3 days total of using flash cards to completely solidify it and ive never forgotten them even when i went years without studying. it uses mnemonics, and i think it can be a good alternative if the brute force method isn't working for you.
3. for other grammar resources, esp if you prefer reading to watching videos, you can try yokubi or the classic tae kim's guide. there's also japanese ammo by lisa for an alternative video series. i don't recommend trying to absorb every grammar point possible before reading- you really just need the absolute basics, and the rest will come naturally as you see them in context while reading hundreds and hundreds of times.
4. for me personally, starting off reading manga for a month or two helped me dip my toes into reading without feeling overwhelmed with literary prose or long dialogue with vns and novels, but it is important to try and graduate to the latter as fast as you can. when i finally stopped being scared to try reading novels i can't tell you how much of a difference it was to enhancing my reading skills- reading actual books, and visual novels too, just blows comics out the water. but! they're still very good for stepping stones. remember your first vn and ESPECIALLY your first novel is going to suck, it's gonna feel like beating your head against the wall. it took me over 30 hours to read my first novel and it was for children lmao. but, every time you finish one it gets easier and easier. anyway, for manga, i recommend the mokuro app there's hundreds of premade ocr'd manga and all you have to do is install them and start reading. it also tracks how many characters you read.
5. game sentence miner + agent agent text hooks games and gsm allows you to use yomitan on the hooked text, it also acts as a repository that tracks your vns and game progress, how many characters you read, and lets you set goals. it even has an overlay you can put directly over your game window so you can use yomitan directly there instead of having to go to your text hook page. gsm also works with lunatranslator. generally you wanna use luna for pc visual novels, and agent for games on consoles. i recommend agent over luna generally, as luna in my experience can be fussy, and agent loads text near instantly to gsm. the switch especially has a treasure trove of visual novels to read.
6. random but do you like tsukihime? there is a 100% browser based version of it with some qol features, and selectable text so you can use yomitan directly in browser with no other resources necessary. you just gotta change the language to japanese in the settings.
it's not necessary to read per say, but i recommend listening as well as reading, even if you don't care about being able to understand verbal japanese in media or never want to speak it irl, it can help reinforce your reading, and figure out the proper pronunciation. you can start with watching shows with subtitles and moving up to pure raw listening.
also yes to op's point it's so important to just read. 99% of any problems you have can be solved with just read more. keep forgetting grammar and vocab? read more. reading is hard and frustrating? read more. consistency is key, and you also wanna try and stick with things that you're interested in, esp things youve already read or watched. i feel like stuff labeled as shoujo or shounen is the sweet spot- complex enough to not be boring, but not so much that it becomes overwhelming.
Text of tweet under the cut because it is loooong.
But... Stochastic Parrots.
Timnit Gebru was fired from Google in December 2020 for refusing to retract a research paper, and every single warning that paper made about large language models has now happened at a scale the industry spent 4 years trying to make people forget about.
Her name is Timnit Gebru.
She co-led the Ethical AI team at Google. She co-wrote a paper called "On the Dangers of Stochastic Parrots" with Emily Bender at the University of Washington and two other researchers. The paper was 14 pages long. It was submitted to a top AI ethics conference. And it was the reason Google decided that one of the most senior Black women in AI research could no longer work there.
The story Google told publicly was that she resigned. The story she told, confirmed by 2,695 of her colleagues in an open letter, was that she was fired by email while on vacation because she refused to either retract the paper or remove her name from it.
The paper had not even been published yet.
Here is what she actually wrote, and why every prediction inside it has now come true.
The first warning was about scale itself. Bender and Gebru argued that training ever-larger models on ever-larger scrapes of the internet would produce systems that appeared fluent but had no actual understanding of language. They called these systems stochastic parrots because they would repeat patterns from training data with statistical confidence and zero comprehension. The paper predicted that this apparent intelligence would fool both users and developers into trusting outputs that were structurally incapable of being reliable.
This was 2020. GPT-3 had just come out. The paper predicted the hallucination problem before anyone had a word for it.
The second warning was about bias amplification. The paper documented in detail that internet-scale training data contains systematic overrepresentation of dominant viewpoints and underrepresentation of marginalized ones. The models would not just absorb this bias. They would amplify it, because the optimization process rewards confident outputs, and confidence in language patterns tracks frequency in the training set.
The prediction was that hiring tools built on these models would discriminate against women. That healthcare triage tools would underperform on Black patients. That loan approval systems would entrench inequality while presenting their decisions as neutral algorithmic judgment.
Every one of those things has now been documented in deployment.
Amazon's hiring algorithm penalized resumes that contained the word "women" in any context. Healthcare risk scoring algorithms used by major US hospitals were found to systematically underestimate the medical needs of Black patients. Apple Card's credit algorithm gave wives credit lines 10x lower than their husbands for the same financial profile.
The third warning was about environmental cost. The paper calculated that training a single large language model produced emissions equivalent to the lifetime output of 5 cars. The prediction was that the race to scale would create an environmental footprint that would eventually rival entire industries.
In 2024, Google's emissions were up 48% from 2019, and the company explicitly blamed AI infrastructure. Microsoft's were up 29%, same reason. Both companies have now quietly abandoned the climate commitments they were publicly celebrating the year Gebru was fired.
The fourth warning was about documentation. The paper argued that the training datasets being assembled were too large for anyone to actually audit. Nobody at Google, OpenAI, Meta, or any other lab could tell you with confidence what was in the data their models were trained on. This was not a temporary problem to be solved later. It was a permanent feature of the approach.
In 2023, researchers discovered that the LAION-5B dataset, used to train Stable Diffusion and other major image models, contained thousands of images of child sexual abuse material. The companies that had trained on the dataset had no way of knowing. The paper predicted that category of failure 3 years before it was found.
The fifth warning was the one Google cared about most.
Bender and Gebru argued that the deployment of these systems would centralize linguistic and cultural power in the hands of the small number of companies that could afford to train them. The internet would become a place where the dominant voice was a statistical average of dominant voices, presented as a neutral assistant. Languages underrepresented in the training data would degrade over time as more web content was generated by these systems and fed back into the next training run.
This is now happening in real time. A 2024 study found that 57% of new web content in English is AI-generated or AI-assisted. Researchers studying low-resource languages have documented active degradation in translation quality, because the synthetic content fed back into training is itself worse in those languages.
The paper Google fired her for predicted the model collapse problem before model collapse had a name.
The mechanism behind why this all happened is the part of her work that nobody quotes.
Gebru's argument was not that AI is dangerous in some abstract sci-fi sense. Her argument was that AI is dangerous in a very specific structural sense. The technology was being built by a small group of researchers who shared similar backgrounds, worked at similar companies, and were rewarded for shipping products faster than competitors. The incentive structure made it impossible for safety, ethics, and bias concerns to slow anything down. Anyone inside the system who raised those concerns was either ignored, sidelined, or removed.
She was making that argument from inside Google.
Then Google proved her right by removing her.
The team Google had built to make sure their AI was safe was dismantled in 90 days because they did the job they had been hired to do. Margaret Mitchell, the other co-lead of the Ethical AI team, was fired two months after Gebru for searching through her own emails for evidence of how Gebru had been treated.
Gebru did not stop. She founded DAIR, the Distributed AI Research Institute, in 2021. The mission is to do AI research outside the control of the companies that have a financial interest in not hearing the answers.
Every prediction in the Stochastic Parrots paper has now been validated by deployment. Hallucinations are an industry-wide problem the largest labs cannot solve. Bias amplification has been documented in hiring, healthcare, lending, and criminal justice. Environmental costs are larger than entire small countries. Training data audits remain impossible. Model collapse is an active research crisis at every major lab.
The question worth sitting with is the one almost no one in the industry will say out loud.
Every researcher with the technical credibility to call out these problems watched what happened to her in December 2020 and made a calculation about their own career. The number of people willing to speak publicly about safety and ethics issues inside the major AI labs collapsed after that firing and has not recovered.
The researcher Google fired for warning about exactly what is now happening was right.
The company that fired her is now the second-largest deployer of the technology she warned about.
And the people inside that company who agree with her are not allowed to say so.
Long before the Sega Hard Girls, ASCII Corporation created two anthropomorphic Saturn mascot characters of their own for their Tech Saturn gaming magazine. The one on the left, representing the Model 1, is Satan-chan. The one on the right, representing the Model 2, is Satan'na-chan.
In addition to appearing in print, both characters featured heavily on the media discs bundled with each volume of the magazine. The link below leads to a video of Satan'na-chan that was ripped from one of these discs:
origamiaround
andulkaphoto
pixelskylines
snowghoul
blakekathryn
