i come in peace

Recently the Guardian ran an op-ed piece written entirely by a neural network, namely OpenAI’s powerful new GPT-3, pitched as a position piece to the readership to assuage as many as possible of the fears that robots would be the destroyers of humanity, employing tools of rhetoric and downplaying movie tropes and even the warnings of Stephen Hawking.
I don’t know if such an appeal should be interpreted as reassuring or not but the litany of supervillains that the same autodidactic subroutine conjured up at the behest of Janelle Shane (previously) is really the antithesis and counter-argument and rather betray the insidious plotting of the machines. One of the more telling characters is Flugg—an evil robot-bunny who is powered by a magical bunny tail—the world would be better off if there were more bunny tails. Check out the whole rogue’s gallery and tag yourself. Klaatu barada nikto.

7x7

bouncing here and there and everywhere: a Finnish maths rocks band—via Things Magazine

wrr-fm: the strange and wonderful account of the first radio station in Texas—via Miss Cellania’s Links

infinity kisses: Carolee Schneemann (*1939 – †2019) experimental montage of her smooching her cats

smashedmouths: an all deep fake rendition of All Star using wav2lip subroutine—via Waxy

the medium is the message: hunting down the first mention of cybersex

eeo: Trump bans diversity training, citing them as divisive, engendering resentment and fundamentally un-American

recessive traits: heredity illustrated with gummy bears

6x6

cassandra drops into verse: a thoroughgoing appreciation of Miss Dorothy Parker (*1893 – †1967)

jazz pigeon: from the same creative studio that asked “Are you tired of being a bird?”—via the Link Pack of Swiss Miss

going postal: the United States may soon see the return of post office offering financial services—see previously

it’s not the heat but the humidity: meta-study suggests that dry air may help the corona virus propagate

the gosling effect: another example of machine pareidolia, wherein a computer detects the Canadian actor’s face in a fold of a curtain—like seeing Jesus in a burrito

susan b. anthony: champion for women’s suffrage rejects Trump’s offer of a pardon for her arrest and fine in 1872 for voting illegally

bredlik

As our artificial intelligencer Janelle Shane (previously) recalls to mind, circa 2016 there was a genre of verse introduced by Sam Garland on observing a cow licking loaves of bread in an unattended bakery and framing the poem from the frame of said cow that enjoyed a memetic moment:

my name is Cow,
and wen its nite,
or wen the moon is shiyning brite,
and all the men haf gon to bed – I stay up late.
I lik the bred.

We had forgotten but just as well as Shane was waiting for the internet attention the style was getting had virtually faded away before training her neural network on the subject to see what it would expound on in the same meter (and the same non-standard Middle English spelling) without undue outside influence. Seeding it with three word prompts (e.g., cow, lick, bread), the neural network created some noble rhymes.

ai claudius

Via Kottke—we are directed to the Roman Emperor Project of Daniel Voshart—Star Trek set designer—who has taken a dataset of over eight hundred sculptures and busts to seed a neural-network to create photo-realistic images of the fifty-four caesars of the Principate, the first period of the Roman Empire that began with the reign of Augustus and ended with the Crisis of the Third Century, which nearly led to its collapse buffeted by civil wars, invasions, economic depression, plague and political instability.
These early days of the Empire were no salad days to be sure but this period prior to the crisis is in contrast to the following one referred to as the Dominate or the despotic phase, beginning with the reign of Diocletian and the downfall of the West. The algorithm was guided and informed by written descriptions in the histories to take into account other physical characteristics in efforts not to flatter or romanticise but show diversity as well as the ravages of rule, age and indulgence. Here is our old friend Claudius, who was rather unexpectedly elevated to the role after his nephew Caligula was assassinated by a conspiracy between senators and the Praetorian Guard. Much more to explore at the links above.

9x9

rewritten by machine on new technology: record industry going after a neural network called Weird A.I. Yankovic that generates parody songs in the style of its namesake—via Slashdot

my beautiful laundrette: elderly couple dress up and model the apparel left in their laundromat—via Nag on the Lake

an atmosphere for simple communication and dating: once Russia cinema reopens, the Ministry of Culture is banning drama and dreary movies until at least the spring of 2021

it’s portraits all the way down: an Inception of self-portraiture—see previously 

search history: a New York Times styles reporter documents and annotates everything term she researched online for a week—via Kottke

be the first to like this post: pigeons look for other career options

the tetris effect: a film about the game’s origins is in production but it won’t be another Battleship—via Miss Cellania’s Links 

karen alert: they keep getting worse

good guy: Billie Eilish’s song Bad Guy performed in major key—see also—via Kottke

upsampling

We’ve seen the built-in bias on display of this neural network application that turned a pixelated image of Barack Obama into an avatar that presents as pretty Caucasian, and Janelle Shane (previously) does a really good job at unpacking what’s going on here with our own tendency for pareidolia codified and amplified.
Not only is the algorithm informed by representation (and under representation) which is highly problematic and is something that the industry desperately needs to redress lest machine learning become the next commercialised embodiment of unreliability, the artificial intelligence delivers what it’s rewarded for delivering, be that a human face or a serviceable suspect that complies well enough with a blurry or grainy image. Thankfully most of the leaders in this sector, faults and all, are taking a pause in sharing their technologies with bad actors, including law enforcement agencies. The application cannot recover details that do not exist—only invent them based on what’s been judged plausible.

aromachology

Having made forays into nearly all aspects of design, Weird Universe brings us the account of how a Brazilian cosmetics company approached IBM Artificial Intelligence Research to commission a pair of complementary, wholly machine-engineered (its collaboration was not completely unheard of but the help was solicited under human supervision for concocting, modelling new blends of existing fragrances).
Absent a robust dataset of aromas at the time, it turned to German fragrance clearing house with some two million formulæ of smell samples from household cleaners to toothpaste flavours and of course analysis of perfumes and colognes to train a program to compose unique inventions—called Philyra, the Thessalian goddess of beauty, healing, writing and perfume, credited with the invention of paper as well as the alphabet also mother to the Centaurs, owing to a visitation from Cronos in the form of a stallion. The neural network, free from human interference created some unique suggestions, resulting in at least two so far being brought to market.

13/10

We selected the same header image as the least cursed one to ease into the ramifications that Janelle Shane (previously here, here, here and here) expertly briefs us on with a preview of the capabilities of OpenAI and how attuned it is to following prompts through this “parody” account it has made of the wholesome Twitter property We Rate Dogs that captures the purpose and tone of the original a bit too well with its introduction and (mostly) generous evaluation. The added element of horror is in the generative gulch (as opposed to uncanny valley) when there’s a glitch in the virtual canine that Shane used to illustrate the ratings for its spoof account but that unease seems to us a distraction from what sort of passable bot armies might be unleashed on any of us contrarians if left unsupervised. Much more to explore at the links above.

by-line or fait accompleat

Via Cory Doctrow’s Pluralistic, we are introduced to the Giant Language Model Test Room created to detect machine learning forgeries, hybrid news items autogenerated and highlights the certain slant, intentional or capitalised upon or not, of predictive text by colour coding the output, the copy of an article based on its own protocols, as an obstacle to targeted content to match targeted advertising—which seriously threaten to undermine education and discourse and reveals what’s written by reporters and what’s writes by machines.

first-past-the-post

Via Imperica, one is invited to build one’s own fantasy Parliament using a Generative Adversarial Network (see previously here and here) to create perfectly plausible virtual members.
Despite inherent bias built in to artificial intelligence that tends to reflect back to us our worse inclinations, I think that these representatives might turn out to be fairly agnostic though there’s no way to gainsay or guarantee that they would turn out any better than the current sitting legislative, apparently willing to squander progress, trust and goodwill by creating one set of rules for the governed and another for the ruling and expect the underclass to gladly accept more austerity and isolation in the bargain. Do let us know how your rotten borough, your pocket constituency fares.

✨#12 – boycotting cheese✨

A family acquaintance has been confined to a hotel and Saudi Arabia, one twitter personality reports, and shares this image of a menu card that strikes me as delightfully pure—first insofar as they would go to such lengths to accommodate Western guests, including at a time like this—during the pandemic, whom was stranded and staying for longer than expected plus through the month of Ramadan.
I also like the level of trust vested in a translating algorithm—since absent anything to check it against, why would one have reason to doubt? Also, interestingly, punctuation seems as important as letters, which seems right in hindsight for someone unfamiliar with the script but had not occurred to me before. That said, tag yourself. Foul, fool, full is probably fūl mudammas—stewed, seasoned fava beans—which is very delicious. We had a hard time choosing between Chicken Dump Truck, A Regular Erika or She is Suspicious of Cheese—and wonder what the story is behind dishes such as Friday, Tuna is a Problem and Worried. Beans, gentlemen.

6x6

why that’s a perfectly cromulent word: neologisms coined, defined and used in a sentence by a machine learning algorithm—via Things magazine

elrodon, son of halcyon: anti-depressant (see also) or Tolkien character—via Super Punch

your perfectly creased coordinated casuals: Kristen Wiig reads the early work of Suzanne Somers—via Nag on the Lake

specious logic: Trump argues against testing and tracing

howards end: E. M. Forster’s prescient 1909 sci-fi foray “The Machine Stops”

the floor is haunted: responsibly confined to our own living rooms, AI Weirdness (previously) imagines escape rooms

blue jeans and bloody tears

Inadvertently creating the new subgenre technofear with a nonsensically subversive and unexpected anarchical message, a team of researchers from Rotterdam—the city that had been slated to host the Eurovision Song Contest (previously here, here and here)—cancelled for this iteration but still held on-line—trained a neural network on more than two hundred of the winners and catchiest entrants from over the decades, generating this number that samples from those common elements. Learn more about the teaching methodology at Ars Technica at the link up top.

torch song

Having encountered this neural network-driven jukebox before in several contexts, we were of course quite impressed but at the same time unable to assay the power the algorithms and machine learning so quite appreciated the developers allowing Janelle Shane (previously) to put a quarter in and demonstrate in an accessible what it’s capable of. If you have ever wanted to know what Baby Shark might have sounded like as performed by the Beatles, then you are in luck or sample the below rendition in the voice and style of Ella Fitzgerald. Much more to explore at AI Weirdness at the link above.

6x6

artbreeder: a fascinating, generative branching experiment that makes unique, derivative art from participant’s choices—via Things Magazine (a lot more to explore there)

may the fourth be with you: a disco tribute to the first film of the franchise (see previously)

topocom: mapping a better tomorrow – a 1971 US Army short

econowives: the trailer for a 1990 adaptation of The Handmaid’s Tale (previously) starring Patricia Quinn, Elizabeth Montgomery, Faye Dunaway and Robert Duvall that’s a strange reverse case of the Mandela Effect (I feel I ought to have known about this yet have no memory of it)—via Messy Nessy Chic

wpa: a look at how the US government funded the arts during the Great Depression

such car: machine learning’s mixed meme metaphors, via Imperica

9x9

horsefly stretches so much time: learning French with these near homonyms that sound like (near) idioms, you know—taon temps tant tends

the lord hardened pharaoh’s heart: as scary as “murder hornets” sound, if they destroy the bees, US agriculture will be in shambles

making muppets: Jim Henson presents a tutorial on creating one’s own puppets in 1969, shortly before the debut of Sesame Street

jukebox: a neural network that’s getting quite good at imitating musical genres and syndicating wholly artificial songs, via Memo of the Air

plastique fantastique: these face shields from Isphere have a certain Avengers’ spy-vibe

do not make me fight you: reminiscent of this montage, stunt choreographer Zoë Bell takes on Hollywood

headspace: cranial collages from Edwige Massart and Xavier Wynn

catamaran: this floating shelter in Amsterdam, de Poezenboot, finds new forever homes for our feline friends

www: this was the internet we were promised—why did it take the collapse of civilisation to bring it?

selfie2waifu

Via Boing Boing—though we’d recommend maybe not experimenting with it on one’s phone as it seems to deploy harmless though slightly irritating adware making it easy to fumble over an errant click—we find this fun little application that utilises unsupervised generative adversarial networks (previously) to transform one’s mugshots into manga renderings that make nice and unexpectedly abstract avatars. In the glossary of anime fandom, a waifu (wife)/husbando is one’s fictional counterpart so was a little unsure of the choice in naming and I’d categorise myself IRL with the trait of アホゲ—that is ahoge or idiot hair, describing unruly hair going in all directions, but it is just something amusing to experiment with. Give it a try and show off your results.

netherstan

Here are some relatively harmless neural network-created fantasy flag mash-ups of the personal ensign of the royal family of Korea combined with the flag of the East African Community or Tonga through the filter of Saint Vincent and the Grenadines, though most outcomes are a bit more dicey and some seem absolutely provocative and bent on igniting world war.
If those aren’t enough to incite at least an international incident, one can use the same data-set and vexillogical protocols that the bot draws from (presumably ignorant what national banners can symbolise for some) to create one’s own remixes. Give it a try and share your best unlikely geopolitical union.

lexus-nexus or in this corner

The always brilliant programmer behind AI Weirdness Janelle Shane’s latest foray into machine mentorship begins with a nice and reflective acknowledgement about one primary, force majure why bloggers blog in the first place: to be introduced to process and jargon outside of field of speciality and indulge in learning something new.
This latest episode (see previously) involved in rem jurisdiction—a concept in legal code that imbues an inanimate object status and agency rather than its minders or responsible parties, a type of legal fiction that has resulted in some preposterous sounding suits such as the United States v. One Book Called Ulysses or the United States v. Four Hundred Twenty-Two Casks of Wine. Given that those actual are precedential cases on the books, Shane wondered what her neural network might glean from studying millions of legal proceedings related to seizure and customs violations to create epic courtroom battles. Quite the courtroom artists, Shane has illustrated some of the less abstract ones like Texas v. One Small Dog with a Napkin Near It—quite a surreal enough judgement—but there were others that far exceeded rendering like South Dakota v. an Apparition at a Shoe Store.