@attacus.net
Computational entomologist, privacy enthusiast, historian. Perpetually raging against the ghost in the machine. Mastodon: @attacus@aus.social π Digital Rights Watch | π Byte Into IT on π» 3RRR FM | π PyCon AU
#langsky
#UtLinguaeVitamHabeant
"β¦fins a la dΓ¨cada de 1970, en el marc dβuna polΓtica lingΓΌΓstica orientada a la promociΓ³ de la llengua estΓ ndard, les altres varietats eren percebudes com a desviacions fonolΓ²giques que calia corregir o fins i tot eradicar."
www.vilaweb.cat/noticies/el-...
ATTACKING DEMONS!πΉβοΈ #medievalsky #medievalart #medieval #middleages #demon #demonio
But the real pattern is more disturbing. Our research found: AI-generated code contains 322% more security vulnerabilities 45% of all AI-generated code has exploitable flaws Junior developers using AI cause damage 4x faster than without it 70% of hiring managers trust AI output more than junior developer code
I keep seeing some version of this when it comes to AI-generated code related stats being shared and I am just begging people to dig deeper. I saw folks on here share a stat that says, "AI-generated code contains 322% more security vulnerabilities." And I'm like okay, what's the source on this? π§΅
A great talk by @attacus.net from this year's #PyConAU
All models are wrong, but some are useful.
Aw, thank you!
My father was supposedly one of the smartest people in the world, and he hated that elitist nonsense.
He loathed the idea that intelligence is something you're born with, a single trait that can be measured. It leads to eugenics and bad policy.
My latest newsletter:
buttondown.com/charliejane/...
I donβt mind people knowing it was me π
In food history, we notice how super-refined foods--gelatin, vienna sausage, white bread--lose class and become horrifying once they become industrialized. Suddenly, fingermarks in bread dough demonstrate skill, not clumsiness.
Post-chatGPT, I'm feeling this about writing. Smoothness feels gross.
Last weekend I gave a talk at #PyConAU.
Itβs about quality, scale, LLMs, science, systems thinking, hype, & what happens when the subjective nature of reality meets the assumption that the human experience can be accurately represented to or by machines.
You should watch it!
youtu.be/GBHGHrlRlKs
Medieval manuscript image of a spotty green creature with a long, trumpet-like nose, two hoofed feet and a short tail.
nosu, f.n: nose. (NO-zuh / ΛnΙ-zΚ)
Image: Bible; France, c. 1270-1290; Bibliothèque d'Agglomération de Saint-Omer, Ms. 5, f. 246r.
#OldEnglish #WOTD
Itβs so beautiful, and learning about it made my entire day
Indeed. Interesting choice, that.
An LLM-generated list of Australian Prime Ministers, including βJohn Gillardβ, two Scott Morrisons, and a lot of spelling and factual errors.
Hereβs the one it generated when I tried.
Maybe the real Ship of Theseus was the Ship of Theseus we made along the way
important context for headlines like this: Altman also runs a company (Worldcoin) that he claims will solve this problem he also created
Melbourne Folks! I will be hosting a book launch to celebrate The Mechanic and the Luddite! Come hear me chat about the book with Lizzie O'Shea and Chris O'Neill β and have a free drink!
β’ August 14th, 6pm, Victorian Trades Hall
β’ Register for free: events.humanitix.com/sadowkski-th...
We should be putting a gargantuan amount of money into solar and storage. War economy levels. It should be free to have a guy from the government solar your house up. Electricity should basically cost nothing.
Oh yeah? Just wait until they hear about the human capacity for relentless perfectionism π
To be fair, the meeting included us investigating the log events generated by us investigating the log eventsβ¦
Sleep-deprived Effy has proven too much for Zoom's meeting summary AI, which has managed to produce something significantly less coherent than even my babbling. ft @attacus.net
David R. Hagen just solved a small mystery that I mentioned 13 years ago in the mouseover text of a comic drhagen.com/blog/the-mis...
Chatbots β LLMs β 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 theyβre βrightβ itβs because correct things are often written down, so those patterns are frequent. Thatβs all.
Art by Frey Catton and ZZ Claybourne. Gray word balloon against a black background, text reading "Maybe I'm not saying anything we don't know, but isn't that all art/communication anyway? Reinforcement of the known to battle the untrue?"
You counter evil with real life.
Handwave it because by then βAI will be good enough to solve that oneβ, the same way it will for the climate π
Recently I found myself on Antipope Instagram, where there are endless reels of the current Antipope of the Palmarian Catholic Church.
This is, of course, extremely my jam, and I started reading up on them hoping to find some wild theology, but instead I found something WAY better...
African workers labeling data and providing content moderation are the invisible backbone of AI development. They perform critical labor under challenging conditions, often for minimal pay, yet remain absent from our conversations about the AI value chain.
Someone should have called Semple an βinhueΓ«ncerβ by now, surely
www.theguardian.com/artanddesign...
Just had a sudden mental image of a melting moment shortbread cookie, but they were called βteachable momentsβ and filled with faux pas and the experience of sticking a fork in a toaster
I don't think that "bikeshedding" is a good name for what it represents, we should discuss alternatives
I was interviewed by @attacus.net for last night's broadcast of Byte into IT on RRR Melbourne Independent Radio. This is such a great conversation about my new book! Honestly one of my favorites so far. Huge thanks to @attacus.net. Listen to the episode here. www.rrr.org.au/explore/prog...