björn
120.1K posts
björn
@bj2rn
working on pretraining and model scaling. now @smallco, prev @bigco. freedom lover. guobaorou enthusiast.
🍊 Katılım Ağustos 2009
972 Takip Edilen560 Takipçiler
@frogs4girls I think London and Paris feel bigger than they are because they’re so dense. Also a lot of diversity from area to area, landmarks spread out across the city and lots of tourists on top of locals, which makes it feel more labyrinthine and busier.
English
i have learned that A: london isn’t even in the top 30 cities in the world by size and B: i have got to go to china
anna !!! :)) 🇵🇸🌸✨🌸🇵🇸@frogs4girls
i do not think i understood how large london is. this is a very large city.
English
This was always true. If you're a language model trained on web text your output distribution will have strong priors for the model being trained by OpenAI since most mentions of LLMs on the web will be talking about OpenAI. There's nothing magical or sinister about it
William Wale@snigus
I have a small language model and it’s been pre trained. Now I post train it to say “I’m a language model”. With no mention of openAI The trained model still ends up saying it’s a LLM made by openAI. Even tho OpenAI is never mentioned in the instruction tuning dataset, and there in fact is one sample that says “I am being developed by Anthropic” (not true)! Makes me think models saying they’re made by so and so is pretty weak evidence of copying /stealing / distillation.
English
I have a small language model and it’s been pre trained. Now I post train it to say “I’m a language model”. With no mention of openAI
The trained model still ends up saying it’s a LLM made by openAI.
Even tho OpenAI is never mentioned in the instruction tuning dataset, and there in fact is one sample that says “I am being developed by Anthropic” (not true)!
Makes me think models saying they’re made by so and so is pretty weak evidence of copying /stealing / distillation.
English
@hishaundi it really isn't unless you're talking about greater london in which case you might as well include 경기도 too. but if you had ever lived in both you'd know this ofc
most of london outside zone 1-2 has fuck all going on anyway. enjoy cricklewood or whatever i guess
English
London is much larger than Seoul btw
björn@bj2rn
felt the opposite moving here from seoul. didn’t expect it to feel so small, a cute little town
English
felt the opposite moving here from seoul. didn’t expect it to feel so small, a cute little town
anna !!! :)) 🇵🇸🌸✨🌸🇵🇸@frogs4girls
i do not think i understood how large london is. this is a very large city.
English
Don't cry because it's over. Smile because it happened.
ʕ•ᴥ•ʔ@GuileWinQuote
Korea Air got Brood War as an inflight entertainment option in case anyone was wondering
English
Too many K-pop songs are fully in English. When does it stop being K-pop and just become generic, formulaic global pop slop?
The irony is by chasing international appeal, K-pop is losing what made it distinct and what hooked fans in the first place.
If it doesn't remain distinctly Korean, it will continue to morph into bland, formulaic, globohomo pop.
EMPIRE@EmpireIX
BTS RM Quote resurfaces, causes concern: "We don't want to change our identity to get to Number One. If we sing in Full English, then that's not BTS' Their New Song “SWIM” is a fully English song, however it's topping all Korea charts.
English
Learning to write kernels might be the highest-ROI activity for displaced SWEs:
→ prereq: reasonable engineering ablity
→ six to twelve months of study
→ millions of dollars, mark zuckerberg showing up at your house to hire you, etc.
i wish this were an exaggeration
English
people are taking this the wrong way. look at it like this:
Composer 1. DeepSeek (probably)
Composer 2. Kimi (certainly)
Composer 3. likely will be the first frontier base model pretrained from scratch for a single domain (coding)
it’s a scary thought isn’t it
Fynn@fynnso
was messing with the OpenAI base URL in Cursor and caught this accounts/anysphere/models/kimi-k2p5-rl-0317-s515-fast so composer 2 is just Kimi K2.5 with RL at least rename the model ID
English
old enough to remember when everyone complained about the change from the sedate “madeleine albright passport” to the giant jingo eagle gw bush administration one
Baudrillard Forever@GroovySciFi
Renewing my US Passport. The old design (on the right) is full of color and civic pride. The new one (left) is just transparently a sad little zogpass.
English
@kipperrii after years of using naver maps i can't help but feel the opposite way. google maps is just sooo slow
English
ok this is beautiful because google is not allowed to map korea. they have domestic apps but naver and kakao maps are both kind of bad. when i'm in korea i just use the stars
Junyoung Seo@jyseo_cv
What if a world model could render not an imagined place, but the actual city? We introduce Seoul World Model, the first world simulation model grounded in a real-world metropolis. TL;DR: We made a world model RAG over millions of street-views. proj: seoul-world-model.github.io
English
@joodalooped this looks so much like the old macos twitter client. miss those times
English
