Θ(log(log(yo))) - [email protected]
banner
logarithmika.bsky.social
Θ(log(log(yo))) - [email protected]
@logarithmika.bsky.social
58 followers 130 following 310 posts
did you know that 29 cats in a trenchcoat can convincingly pretend to do mathematics? 33, he/him, Chilean fan of: cats, birds, Touhou, SCP Foundation, Mother3 (&related), Celeste interests: math, programming, writing, videogame music might have opinions
Posts Media Videos Starter Packs
listening to the giygas boss battle music to concentrate while reviewing something
Reposted by Θ(log(log(yo))) - [email protected]
whiny baby man isn't satisfied having more money than everyone else combined. mad about factual depictions of him in some books, pays to copy every single one and rewrites just those to put in a new building he tells everyone to go to instead

also why he bought twitter, and the entire point of grok
it's super clear that the reason every rich guy and company wants people to ask their genAI all their questions is so they can easily intercept and insert answers for questions that benefit them or whoever will pay them. it's a shortcut to how Google abused the position it took them a decade to gain
for some reason grok is responding in first person from elon musks perspective lmfao
mega starmie is the embodiment of the "threatening images with earthbound music" meme
Reposted by Θ(log(log(yo))) - [email protected]
TLDR; The PSF has made the decision to put our community and our shared diversity, equity, and inclusion values ahead of seeking $1.5M in new revenue. Please read and share. pyfound.blogspot.com/2025/10/NSF-...
🧵
The official home of the Python Programming Language
www.python.org
(Funnily enough, I did test my exercise with small numbers, but I *had* to choose just the few combinations that produced exact answers. I have a guess why, but that's off topic, I suppose)
So yeah, once again no actual reasoning going there. I still find them somewhat impressive as a language model, but checking your answer with small numbers is something virtually all students do, and yet the AI didn't because it wasn't explicitly asked to.
For some detail: it was a counting problem about collars and beads, so the answer should always be an integer; however, with four beads the expression gave an answer of 3/2 instead of 2, which is entirely nonsensical. Really easy to check, but of course the LLM just followed instructions as expected
Next day I realized there was an error in my intended solution and the problem couldn't be solved that way. Welp. I had to replace the question with something else. Then I realized that any moderately cautious student would have found the issue easily, by testing with actual numbers
so chatgpt gave me more or less the answer I was expecting, with some bullshit reasoning that seemed somewhat circular but indeed arrived to the final algebraic expression I intended— all under expected parameters so far. However...
on the topic of LLMs, I had a really funny experience a couple days ago: I decided to upload the homework I gave to my students to chatgpt to see what kind of answers I could expect from someone using AI to cheat
Reposted by Θ(log(log(yo))) - [email protected]
I cannot tell you how many tech journalists at prominent media organizations do not understand this
Chatbots — LLMs — do not know facts and are not designed to be able to accurately answer factual questions. They are designed to find and mimic patterns of words, probabilistically. When they’re “right” it’s because correct things are often written down, so those patterns are frequent. That’s all.
so, #300 posts already. haven't made many drawings yet so here's some fanart of The Amazing Ergodic Pig! cosplaying as the math goblins from SMBC

also you should become aware of The Amazing Ergodic Pig! as they have very strong (and objectively correct) mathematical opinions
thinking of that brief moment in time where len'en was real
"featuring Dante from the Too Bad I Already Portrayed You As The Wojak And Me As The Chad series"
Reposted by Θ(log(log(yo))) - [email protected]
only just realized the wikipedia article still calls it twitter. truly the worst rebrand of all time
there are two queues, one of them just has a fancy logo with three arrows
ok definitely my favorite 9kcon entry (so far)

logging in just to upvote this one
reading the scp-9000 about an extradimensional queue while waiting in a super long queue is certainly an experience
Reposted by Θ(log(log(yo))) - [email protected]
while I acknowledge that LLMs do have some uses for stuff like “producing code with knowledgeable human supervision,” this perfectly reflects how I feel about using it for my own creative work
David Simon, creator of ‘The Wire’, being interviewed by Ari Shapiro (NPR)
acá yo preguntándome por qué "rené Goscinny" estaba escrito con minúscula por un segundo hasta que me di cuenta de que la r es por rue

ni sé si es por ese Goscinny de hecho
mi lado fan de las matemáticas discretas me pide más viajes en metro
Reposted by Θ(log(log(yo))) - [email protected]
as a girl with a PhD in natural language processing and machine learning it's actually offensive to me when you say "we don't know how LLMs work so they might be conscious"

I didn't spend 10 years in mines of academia to be told ignorance is morally equal knowledge.

We know exactly how LLMs work.