ponder

@ponder.ooo

3324 Followers 248 Following

fractals georg

osculating torus
clouds and such
modest quantities of gunpowder
cat in the corner of a creeper farm
"ai concierge" thoughtlessly pattern-matches my issue onto a predefined class of issue that is not applicable and tells me it's my responsibility when it's not
the condition that the jacobian determinant is non zero is equivalent to f being locally invertible
some basic ass sandboxing
this week, we published a blog on improving safety and alignment
benchmarks run in a highly isolated environment, with network access constrained to the ability to install packages through an internally hosted third-party software that acts as a proxy and cache for package registries.
a highly isolated environment, with network access
git readme "just running the init script will create privelege escalation vulns on ur user account"
crap render
tops of the letters are cut off
big bird
gparted sees half the drive now properly unallocated
dire situation in gparted
this is not a place of honor
unsafe code yay
glm suggests patching ntfsresize to ignore the bitmap check
the short and hard truth is: you cannot fix $Bitmap cluster accounting mismatches using native linux tools
quote of ball describing his dystopian hellscape scenario
nilered extra has bought another huge machine
guy describes the trump admin directing every agency to issue "soft law that creates FUD", notes that "it needn't be that well justified"
full AI communism, AI is a "public good", "digital public infrastructure" the author of the paraphrased passage considers this a "dystopian hellscape" and says he's "never met an open-weight models advocate who doesn't ultimately concede this is where things end"
open-weight models are inherently decelerationist
i am surprised china is fine with it. i suspect the reason they are is 75% explained by strategic blindness/lack of AGI-pilledness (the CCP is very Yann Lecun-y in its views of AI)
specimina
noisy fractal
additional fractal
fractal
log blocks the path
objects
quarry
diagram relating text in a paper to the figures it references. everything is all over the place
stochastic
exact average across sequence offsets
fractal
full image, philosophers arguing
👆
screenshot of the page i am posting this message from. it's all wingdings
fractal
the real issue was that the browser hadn't been served the fresh css T_T sorry i got snappy w/ u abt it uwu
please assume i have correctly reported the behavior i am seeing and re-evaluate from the position of believing the issue i am running into is real and is as i have described it
fractal
jormungandr but he's got his party hat on
impro.social but there's a border between feed entries and the line tying replies to their parent post is slightly brighter. also interaction counts have been obliterated
dog hiding under my desk
the file is about 195kb. let me read it
ajj guy
a ponder in red lighting wearing band merch like some kind of ajj shill
writing on the sidewalk including "pcc slab"
With the escaping air and pressure, gross dismemberment ensued; it included bisection of his thoracoabdominal cavity, which resulted in fragmentation of his body, followed by expulsion of all of the internal organs of his chest and abdomen, except the trachea and a section of small intestine. These were projected some distance from the bell, with one section being found 10 metres (30 ft) vertically above the exterior pressure door.
Large language models do not scale logarithmically. When they’re given more words to process, they get slower and use more memory—the time and resources increase faster as the input grows. In technical terms, LLMs scale quadratically. Any computer-science student knows that this is very bad.
Ilya Sutskever, a co-founder and former chief scientist at OpenAI, said in a November interview that companies take the brute-force approach “because it gives you a very low-risk way of investing your resources.” It’s harder, he argued, to invest in research that would reengineer a product currently accruing trillion-dollar valuations.
Yet the returns are diminishing. The bigger an AI model is, the less it improves with each added parameter, and so it must be made bigger at a faster rate just to sustain steady progress. I asked a few AI researchers whether they could name any other real-world software that scales so poorly. None of them could think of any.
the industry’s observation that bigger models tend to outperform smaller ones has given rise to a totemic belief in “scaling laws” that suggest any problem can be solved by simply making models bigger