Replying to on social.treehouse.systems
again I want to reiterate -- the global economy depends on a distro where 56% did not elect to participate in the questionnaire concerning the largest economic issue of our time
Remote
Ariadne, bunyote friend 
@ariadne@social.treehouse.systems
mastodon 4.6.7+glitch-tha feral indigenous (keetoowah cherokee band) cryptid living on duwamish land and building systems that aspire to the quiet empathy of the forests she wanders on weekends...
fediverse is my place to be unfiltered, if all you care about is tech, find me on *groans* linkedin
(also, in case it wasn't obvious, my opinions are mine alone and do not represent those of my company or any other entities relating to me)
7825 Followers
890 Following
50 Posts
Joined April 05, 2022
Website:
Pronouns:
she/her or they/them
Replying to
and if we go inside the sphere and also add a wireframe... #ludereposting
RE: https://sfba.social/@drahardja/117341398329826875
from a climate perspective, a moratorium on commercial AI is *very* important because they are burning fossil fuels *on site* at these datacenters because the grid cannot deliver enough energy.
Quoting
It’s galling that tech billionaires tell us that letting them destroy the climate to develop AI will allow AI to tell us (somehow? maybe? probably?) how to fix climate change. And even more frustrating that news outlets report this blather without so much as a wrinkled eyebrow.
MY DUDES WE ALREADY KNOW HOW TO FIX CLIMATE CHANGE. We stop spewing carbon into the air; install solar, wind, hydro, geothermal, and other renewable energy generation; build mass amounts of efficient public transit; and stop funding industries that dig fossil fuels out of the ground.
AND WE STOP DOING AI.
We already know how to fix climate change, and we aren’t doing it. Even if some superintelligent chatbot COULD figure out how to fix the problem after we turn our world into an absolute oven, it’s bold to assume we would listen to good advice then, because we (including techbros, ESPECIALLY techbros) aren’t even listening to good advice TODAY.
These people are tedious liars and scam artists.
#ai #scam #climateChange
Open quoted posti write code quickly because i write less code.
she said yes 😭
Boosted by @dansup@mastodon.social
Replying to
@ska@social.treehouse.systems I only rep #pixelfed by my man @dansup@mastodon.social
miss me with that insta bullshit
Replying to
i hope i don't seem too egotistical, i just think the webcam input is fun x_x
Boosted by @trending@homestead.social
look if you are going to send us a report complaining that someone is making fun of charlie kirk, we are just going to assume you're a nazi at this point
fuck offfffffffff
Replying to
and this is where dario's new "pacing" narrative becomes useful.
if frontier development is going to become naturally gated by hardware generations and datacenter buildout anyway, then a slower release cadence no longer has to look like the scaling story is running out of steam.
instead, you can frame that cadence as intentional: we are slowing down because we are "responsible", because we need time for evaluations and safety work, because the frontier must be paced.
that is a much better story to tell investors, who are eagerly waiting for a credible path to ROI at this point, than "the next meaningful scaling step has to wait for another generation of accelerators and a few gigawatts of infrastructure."
Replying to
the scaling laws themselves will likely continue to hold just fine, but the bottleneck is increasingly the infrastructure required to realize the next point on the scaling curve.
for a concrete example: a 1-trillion-parameter dense model at 16-bit precision needs about 2 TB of memory just to hold its weights. that's roughly 25 nvidia H100 accelerators just to load one instance of the model.
and once a single inference has to cross dozens of accelerators, memory bandwidth and chip-to-chip communication become part of the latency budget too.
at that point, “make it bigger” is no longer mainly a model-design problem. it is a hardware, networking, power, and datacenter problem.
we've tried nothing and we're all out of ideas
Replying to
when will they invent bunny HRT
I was today years old when I heard the Muzak version of Smells Like Teen Spirit and let me tell you this didn't need to exist
Replying to
and the timing for this essay drop matters here: anthropic is preparing to go public.
an IPO means asking public-market investors to underwrite a company whose future value depends heavily on the assumption that frontier capabilities will continue advancing, while the cost and physical difficulty of producing each new generation keeps increasing.
if the actual development cadence is about to become more visibly constrained by hardware and infrastructure, you need a story for why that is happening.
"we have reached the point where responsible AI development requires us to pace the frontier" is an extremely good story, even if the real reason for pacing is due to physics: they simply cannot build the necessary infrastructure at the current scaling pace.
Boosted by @dansup@mastodon.social
Replying to
@Elizafox@social.treehouse.systems i would not believe fedi gossip, half these people are just pieces of shit looking for clout by starting shit.
Replying to
as an experiment, i decided to implement mastodon's blurhash algorithm against the album art to generate accents for the playing page
i think it came out pretty well...
Replying to
the growth in infrastructure requirements just to scale models dramatically changes what frontier scaling looks like over time.
instead of a relatively smooth progression where each new generation mostly waits on another training run, you get something much more stair-stepped similar to the intel tick-tock model:
new accelerator generation (tick) → build out enough capacity → scale the model → optimize around the new bottlenecks (tock) → plateau → wait for the next hardware generation.
in other words, the pace of frontier capability starts getting set less by model architecture and more by nvidia, tsmc, hbm supply, networking, power delivery and datacenter construction.
Eurostar wanted 300 euro so I guess we're doing this the hard way. #NowTräwelling
https://traewelling.de/status/9135506
anyway, i've had a few people ask for my take on dario amodei's "we must pace the frontier" essay, so i am going to give it a quick read and then write a rebuttal
Replying to
anyway if you do not like $policy, then what you do is quietly circulate your resume and find a new job where $policy is not a thing
Replying to
ok, figured out how to tell QML that i want to render the album art as hidpi :)
Replying to
in light mode, the blurhash effect is not as intense as it could be. time to play with the CSS.
Replying to
again, demoscene roots and all of that
Replying to
maybe something like this, but i suspect the real answer is to have the controls float somehow.
again: i don't know what i am doing here, i am just designing and throwing away shit i don't like
Replying to
this is how audacious has always been built btw, we write some code, try it out, see if its good or if its shit
Replying to
(i also think using chroma-sampling like blurhash does is a nice callback to audacious' demoscene roots)
Replying to
we are incidentally assembling this plane while it is already flying
Replying to
https://github.com/audacious-media-player/audacious-plugins/pull/214 if you want to try it
Replying to
i like this blurhash thing so much i might also apply it to qtui
Replying to
i don't think i am digging the top bar thing that i copied from the kirigami tutorial i am following
I will begrudgingly concede that Germany's obsession with vending machine stores ("e-kiosk stores") is neat
Replying to
yo come to FREENET we got DSL
is the DSL free? because that is what DSL is worth in 2026.
Replying to
@mnl@hachyderm.io @immibis@social.immibis.com @tante@tldr.nettime.org
they should be able to use their silicon on their own terms.
and that is the point.
the billionaires who control these AI systems don't want that for you, they want you to use *their* silicon.
it's feudalism all over again.
and, obviously, feudalism isn't punk.
Replying to
@jakob@social.jakob.space maybe i will respond asking about other windows apps which use this SDK
do you go woof in slack (or whatever other work chat you use)? #AriadnePoll
do you go woof in slack (or whatever other work chat you use)? #AriadnePoll
Replying to on social.treehouse.systems
(i need to figure out the hidpi issue with the album art)
i'm glad digital ocean banned me years ago because i created a bot that automatically opened a ticket every time packet loss exceeded 35%, really saving me time now
Replying to
@mnl@hachyderm.io @immibis@social.immibis.com @tante@tldr.nettime.org
i mean, gemini is more punk than the web, but sometimes pragmatism wins
i'll admit it, at this point in my life i'm a poser
Replying to
@dansup@mastodon.social but why make an ad
Replying to
@dansup@mastodon.social real question: how do you expect to hit that target if latest is constantly moving
Replying to
@dansup@mastodon.social i did not need to know that shrek drag queens existed, thanks
Replying to
like I want to reiterate, DSL is BAD.
it is about fiber, docsis 3.1+ and 5G man.
and they say the US is backwards
what is #ludere capable of? well... this.