Home | Notifications | New Note | Local | Federated | Search | Logout

talya (she/her) but even more queer💄🏳️‍⚧️🏳️‍🌈:jewish_anarchism:@talya@dybbuk.club

READ BEFORE REQUESTING A FOLLOW

🔞 minors do not follow (i don't hate you there's just explicit lewd content in this account. go follow my other accounts instead.)

this is your friendly fedineighbour talya. you might know me as @Yuvalne or @yuvalne or @yuvalne or @yuvalne or maybe even as @talya (dang I have so many profiles). this is the profile where i'm being way more freaky while enjoying GoToSocial. expect swearing, masks to be off, and also other things to be taken off if you know what i mean.

bio copied from @Yuvalne in case you don't know me:
#trans, #lesbian, brown, #jewish (both Ashkenazi and Mizrahi, thanks for asking!), #nerd, classical #musician, somewhat techie, kinda cute. i'm basically Wendy Carlos but significantly less cool. will swear in toots.

proud financial supporter of so many projects I can't list them here (weird flex, I know). see this list for all of them: https://pad.envs.net/pad/#/2/pad/view/l3826rsf9nVcQNz7ihRLo3QA0U+W76NpAOpTem5rGr0/

interact before you FR.

ask me about #TokiPona!

languages i speak: Hebrew, English, toki pona: fluent; Czech: officially B2! Yiddish, French, Arabic: will probably go back to those one day
all my other accounts: https://keyoxide.org/aspe:keyoxide.org:RUGUW6K7H2SMTY2MFH4FE6QNQM
my music: https://negaivevoid.art/talya
pronouns.cc: https://pronouns.cc/@talya
Joined: 2026-09-16 20:34:12
1 notes, 0 following, 0 followers

talya (she/her) but even more queer💄🏳️‍⚧️🏳️‍🌈:jewish_anarchism:@talya@dybbuk.club (2026-09-16 18:54:19)
don't learn languages on linguardia.com.

when i saw its creator mention there's hundreds of languages on it, some of which don't have many learning resources, i already knew there's genAI involved. and while they try to obscure that information, the site is indeed ridden with genAI content. their basic concept is taking translated pairs and analyzing them for grammatical patterns, but the thing is doing something like that with just a pure set of translated pairs and no additional structure requires *a ton* of data, which linguardia doesn't have.
but it actually gets worse. it's not that they went the path many projects go of using biblical corpuses as their pair source (which introduces its own sort of bias), rather they went with Tatoeba (which is a fine-ish source but already not good enough for me to trust it for automated analysis at scale), and then added on top of that *Google Translate* as well as **their own small translation model**. in fact the majority of pairs are machine translations.

all of that's enough of a red flag to rule them out immediately. but just in case, i checked the trial lessons they offer both for Hebrew and for Czech.
even being the most generous i can in scoring when analysis is correct and when it's wrong, the error rate is almost 40%, just for grammar analysis.
but the romanisation and TTS... those are wrong more often than they are right! even for Czech, which unlike Hebrew is not written with an abjad which should make TTS easier, there were serious mistakes.

there's a reason you learn languages with fellow humans. there's a reason Duolingo got worse when they introduced AI, not better.

don't learn languages on linguardia.com.

#Languages #LanguageLearning #Linguardia #lingonardi  #fuckAI #fuckGenAI #AI #GenAI #LLM #LLMs