Mastodon Feed: Posts

Boosted by krig@goto.liten.app:
jessie ("Jess Rose") wrote:

Getting way more use out of this image than I could have imagined back when a younger and more hopeful version of me saved it to my phone.

Boosted by brib@bribstodon.xyz ("brib :neofox_floof:​ :Nonbinary:"):
dinobrarian ("Charlie") wrote:

Ok, stage 1 of the #Quiltagon is complete: I've got a solid fabric layer, with the planets all in place.

Still to do:

- screen (and bits to go onto it - awaiting material)
- LED panels for behind the planets (awaiting materials because I'm NOT using steel bone for that)
- Sequins
-masking layer to hide the LEDs
- batting layer
- backing

However this is now recognisably a full #Tildagon and I've got more than a snowball's chance of finishing this thing before #EMFCamp 2028 (sequins take time!)

ireneista@irenes.space ("Irenes (many)") wrote:

we reserve the word "magic", in software, only for the most deeply incomprehensible stuff. anyway we called this feature "magic comments". we're really quite pleasantly surprised that we managed to write down our understanding of it at all, and we make no apology whatsoever for the overly conversational tone of the docs.

Boosted by glyph ("Glyph"):
AngloPeranakan@lingo.lol ("Guy Emerson") wrote:

This Is Just To Say

I have trained on
the proofs
that were on
our servers

and which
you were probably
saving
for publication.

Forgive me
I can't rule out
that your proofs
improved our models

but burning $15m
on a $1m prize
so lean
and so slop

glyph ("Glyph") wrote:

RE: https://infosec.exchange/@0xabad1dea/117239484057326836

This stuff is really starting to make me _angry_. Every one of these new "achievements" looks plausible and like Lucy with the football, people I trust keep telling me it actually works and surely this time it's real. And then like clockwork, 3-6 months later, it turns out that the efficacy of this new technique is somewhere between "break even" and "fraud".

They're turning our whole industry, my entire life's work, into a goddamn memecoin pump and dump scam and I just don't know what to do.

ireneista@irenes.space ("Irenes (many)") wrote:

that really-for-sure needed documenting, it was the kind of thing which would be way too hard to modify later otherwise

ireneista@irenes.space ("Irenes (many)") wrote:

okay, so that was QUITE the documentation undertaking (we spent, like, just about three hours on it)

but we're very pleased with the resulting write-up

scroll through https://code.irenes.space/evocation/commit/?id=880b631f268d70cf0282203f001afc1f0cc7bf31 until you hit the big block of green text

Boosted by glyph ("Glyph"):
0xabad1dea@infosec.exchange ("abadidea") wrote:

Project Glasswing:

Claiming to have found 26 thousand real vulnerabilities but only 0.8% of them have resulted in a real fix in a real project after five months is dire. They blame it on the human independent review bottleneck, but human experts being paid for their time definitely have a higher throughput than that when working with data that’s actually actionable.

The assigned-at-Claude severity ratings are also dire. It assigns “high” or “critical” to 91% of findings. Most findings in the real world are low or medium. This should be especially true when using a magic machine to shake out every last little issue that was overlooked by humans focused on the biggest risks.

Together this implies it’s generating thousands of trivial or nonsensical findings and labeling them HIGH DANGER CRITICAL MUST FIX, and the human independent verifiers are sifting for the rare needle in this haystack worth passing on. This isn’t really an improvement over the high-noise automated scanners we already had

(This is a corporate blog of someone with their own vulnerability management services to sell, so apply an appropriate number of grains of salt to their analysis. Filter keywords: AI LLM Anthropic)

https://www.vulncheck.com/blog/anthropic-glasswing-receipts

Boosted by brib@bribstodon.xyz ("brib :neofox_floof:​ :Nonbinary:"):
donni ("donni saphire") wrote:

I’m sleepy, but sure, I could go for several more hours of spiraling thoughts

Boosted by glyph ("Glyph"):
jalefkowit@hachyderm.io ("Jason Lefkowitz") wrote:

You think Copilot makes people more productive, imagine how they would fly through work if Microsoft removed all the bullshit from Windows and Office

glyph ("Glyph") wrote:

East Bay friends: if I wanted to meet up with somebody with a couple of laptops and hang out and write code in the evening, somewhere in the vicinity of the southern bits of Oakland, Jack London, or Alameda, where would I go? When I would do this in Austin back in the day there were plenty of waterfront cafés open to midnight but I do not know of anything that goes past - or even really close to - sunset here.

(It doesn't have to be a coffee shop, that just seemed the obvious type of place)

Boosted by glyph ("Glyph"):
SnoopJ@hachyderm.io wrote:

RE: https://neuromatch.social/@jonny/117237800074111038

A recurring thought through all this insanity is "what if we just spent a fraction of that money on [the obvious way] instead" and it always seems like the economics would be way less shitty, so

Boosted by neatnik@social.lol ("Neatnik :prami:"):
mbjones@social.lol ("Brandon") wrote:

just posted on the blog: "what are we doing here"

this is a followup to my last post because i've got thoughts on the belief that its a dogpile to critique a public post these days.

https://krrd.ing/what-are-we-doing-here/

#Blog #Accountability #Faith

glyph ("Glyph") wrote:

I just cannot physically keep up with all the news. Is there some movement on an IPO or something like that which are panicking sales reps that is resulting in cascading knock-on effects to increase the pressure everywhere?

glyph ("Glyph") wrote:

my suspicions that something weird is going on are heightened (if you do not know him, Chris is an ex-NFL player and Democratic politician; certainly not a fan of AI but also not like primarily an anti-AI tech person)

https://bsky.app/profile/did:plc:oofa3qqoiszsmajbigfqskqv/post/3mv2lwfozwc2i

neatnik@social.lol ("Neatnik :prami:") wrote:

"not Polish" 🤔
"foul rap music" 🤔

This is Vincent Ritter, who in addition to these thinly-veiled racist remarks also once posted "When a game asks for 'pronouns' and doesn’t let you skip that part. Uninstall. Sigh." And posted about how sad he was to see the heteronormative Apple emoji being replaced by inclusive, genderless versions.

He pays for a blue checkmark on X, cheers for Elon, and is a member of the Micro.blog team.

If you use Tinylytics or Scribbles, this is what you're supporting.

Boosted by glyph ("Glyph"):
jaafar@hachyderm.io ("Jeff Trull") wrote:

Thought experiment: if I were aware of an error in a paper published in an IEEE journal could I get an AI model to repeat that error or a similar one? Or draw erroneous conclusions based on it?

Boosted by glyph ("Glyph"):
kf@666.glitchwit.ch wrote:

padme MARRIES anakin?

only a man could have written this

Boosted by glyph ("Glyph"):
scream@bots.robots.rodeo ("Endless Screaming") wrote:

AAAAAAAAAAAAAAAHHHHHHHHHHHHHH

jonny@neuromatch.social ("jonny (nonvenomous)") wrote:

In case I didn't caveat enough in the above posts, this is me saying again these are extremely crude back of the envelope estimates with available information and the error bars should be considered wide.

Boosted by glyph ("Glyph"):
teotwaki@mastodon.online ("Sebastian Lauwers") wrote:

No, it’s because they’ve externalised the thinking and have to spend so much energy catching up that it is *physically uncomfortable*, so they start arguing in bad faith. Instead of being collaborative and wanting to improve the outcome, your criticism creates work for them. A person defending LLM output will reach for ad hominems, fallacies and nitpicks to attack the form of your criticism. One you notice it, it’s shocking. (2/3)
@tshirtman @glyph

glyph ("Glyph") wrote:

Has the AI situation gotten markedly worse for you this week?

Specifically: has something happened in your personal or work life, like a mandate to use LLMs at work, or a relative developing an obvious chatbot dependency, to make your own situation worse?

Or have you learned new information from a public source, like social media or a newspaper, that you didn't previously know which is worse?

Boosts for reach appreciated.

ireneista@irenes.space ("Irenes (many)") wrote:

oh wow we found a subtle semantic issue in the layers of metadata transformation that the thing goes through. we know how to fix it, but it's going to be really hard to document the steps in a way that would allow somebody else to fix it...

Boosted by soatok@furry.engineer ("Soatok Dreamseeker"):
gsuberland@chaos.social ("Graham Sutherland / Polynomial") wrote:

are so many network engineers night owls because they're NOCturnal?

Boosted by cstanhope@social.coop ("Your weary 'net denizen"):
raven@fedi.raventhemaker.com ("Carl C") wrote:

I just watched the entirety of a 40-minute video by virtuoso beatboxer Wing of South Korea, explaining his thinking behind developing his recent work _and explaining how to do it_. I'm no beatboxer, I have no interest in even trying, but this young guy sharing his techniques with his competitors (lots of competition in the scene) is just wonderful. His approach to musicality and trying to push beatboxing as a whole forward is how the whole world should work.

https://www.youtube.com/watch?v=qlrpeYdm9Ec

ireneista@irenes.space ("Irenes (many)") wrote:

(after that it'll be time to make it work on proper Forth programs)

ireneista@irenes.space ("Irenes (many)") wrote:

now we just need to add a couple other types of literal, and one more comment formatting feature, and that may be all we need to do for running the hex transform on Evocation-assembly programs

ireneista@irenes.space ("Irenes (many)") wrote:

yessssssss we have Evocation's hex transform including integer literals for values measured at the compiler's "runtime" in appropriate places in the disassembly comments

https://code.irenes.space/evocation/commit/?id=e03216ba8bbc7547b1451ed99497c0693446983b

jonny@neuromatch.social ("jonny (nonvenomous)") wrote:

let's do some more back of the envelope estimates.

no company publishes the energy usage of their models, so as before we just have to do the best we can by making conservative extrapolations.

altman claimed in 2025 that ChatGPT uses 0.34Wh per prompt (this is lower than independent estimates, but let's go with it to be conservative).

Previously i had made the very conservative estimate that considering the "plain info" and retail use of the product bringing it down, the average prompt might be something like 3 paragraphs, or 300 tokens. Later models that use reasoning tokens use orders of magnitude more tokens (prompting claude code to estimate the average number of output tokens without writing code just now used ~2k). At the time of that blog post, the newest general purpose model would have been GPT-4.5, but the consumer model that would have made up the bulk of the average for the power estimate would have been the infamous GPT-4o, a ~200B parameter model - neither of these are "reasoning" models, so lets keep our 300 token estimate for pre-"reasoning" models.

So that would be 0.34 / 300 = 0.00113 Wh/token.

According to ~ rumors and speculation ~, people believe that GPT-6 astra is a ~5T model, and that this internal model might be ~10T parameters. Parameter count is a very imprecise estimate of energy usage, because energy usage depends on a ton of things like hardware efficiency, the attention mechanism selecting how many of the params are active, and so on. But because we don't have any better information, we assume that energy usage is quasi-linear with number of parameters.

So then that would be (10 trillion / 200 billion) * (0.34 Wh / 300 tokens) = 0.0566Wh/token for the giant 10T internal model.

Then multiplying that through for 300 billion tokens yields us 17 GWh. That's roughly a third of LA county's daily residential energy use, a county of 10 million people.

Boosted by neatnik@social.lol ("Neatnik :prami:"):
numist@xoxo.zone ("Scott Perry") wrote:

RE: https://social.lol/@neatnik/117227874076649666

"I have almost reached the regrettable conclusion that the Negro's great stumbling block in his stride toward freedom is not the White Citizen's Counciler or the Ku Klux Klanner, but the white moderate, who is more devoted to "order" than to justice; who prefers a negative peace which is the absence of tension to a positive peace which is the presence of justice; who constantly says: "I agree with you in the goal you seek, but I cannot agree with your methods of direct action"; who …"