What do you get if you instruct an AI to turn a house into the most haunted house in the world? What if you ask it for the LEAST haunted house? How does an AI know what "haunted" looks like, anyways?
I did some experiments with CLIP+VQGAN (link and tutorial here). This makes use of CLIP, which learned what kinds of images and text are associated with one another on the internet. Since CLIP has probably seen lots of things labeled "haunted", it can probably guide VQGAN into making some seriously spooky alterations to this lovely 1894 Victorian house:
For my first try, I had it start with this image and alter it to look more like "A haunted Victorian house".
A haunted Victorian house
The wood has turned to stone, the yard has proliferated more iron fences (and become greener), and one of the turrets has changed to a chimney pot. Other than that, it looks pretty ordinary.
I decided to kick it up a notch, and told the model to head toward "An extremely haunted Victorian house | dramatic atmospheric night photography."
The walls became wobblier, another of the turrets changed to a chimney, and curiously, there now seems to be a smaller more modern house inside the larger house.
It turns out a lot of this is the "dramatic atmospheric night photography".
Left: A victorian house, dramatic atmospheric night photography. Right: A normal very not haunted victorian house, dramatic atmospheric night photography.
But the interface I'm using lets me assign negative weights to things, so I can actually subtract "haunted" from the house I'm trying to make. It sort of works! Once I subtract enough "haunted."
Specifically, this is minus two "haunted"
Apparently CLIP has learned that if you want to make things less haunted, add flowers, street lights, and display counters full of snacks.
This, in case you're interested, is the least haunted place in the world. It's what you get when you give CLIP+VQGAN no instructions except for a negative penalty to "haunted".
I'm not sure what most of this stuff is supposed to be, but if you can find it and bring it to the middle of a baseball diamond in an office park, you will be safe.
I also tried haunting a Waffle House.
"World's most haunted waffle house", starting from this image by MBrickn.(cc-by-SA)
Become an AI Weirdness supporter to get bonus content: a haunted bookstore, a haunted bakery, and a haunted grocery store. Or become a free subscriber to get new AI Weirdness posts in your inbox.
Imagination is a uniquely human ability to visualise unlimited possibilities starting with a simple question: What If. Here, creative studio Six N. Five raise their voices to wonder about the next steps of humankind and a possible future move.
(back of a quirky literary novel voice): Sometimes, things are not what they seem. An architecture critic disappears for three months to follow bike racing around Europe, rife with questions of becoming and desire. A real estate agent uploads a listing to an aggregator, knowing that it will be a difficult sell but thinking not much of it, for, like Tolstoy’s unhappy families, all houses are difficult to sell in their own way. A house is built in 1980 in Staten Island and would have thrived as an anonymous bastion of tastelessness had the internet not been invented. But the internet had been invented. All of these things are brought together here, through truly unlikely circumstances.
Let’s not bother with the formalities this time.
None of you will buy this house.
Sitting Room
Does anything here make sense? The periwinkle sofa, the twinkling of bronze glass, a truly transitional material, a mall exiting stagflation and entering the sultry trap of Reaganite libertarianism that would leave it empty twenty-five years later. The sense that one is always changing levels, trapped in a landing of some sort, never quite arrived on stable footing. But that’s just the style, one assumes. One foot in the seventies, with all their strife, one foot in the beginning of what felt like the end of history. One’s ass on the iridescent pleather sofa, waiting for the centuries to change.
Sitting Room II
My suspicion is that there are no pictures of the mirrored mystery foyer because the photographer’s identity would be henceforth revealed, and the point of all real estate photography is for the viewer to imagine themselves as the only person in a given space.
Dining Room
The shinier things are, the richer one is, obviously.
Kitchen
This serious sociological research also happens to coincide with the Giro d'Italia, one hopes.
Landing
(crediting @cocainedecor on twitter for their term. but also, where can i get some chevron mirrors, asking for a friend.)
Master Bedroom
just asking questions
Bedroom 2
Ostensibly bad opinion that I will nevertheless defend: the corner bed slaps, let’s bring it back.
Basement
(Staten Island accent): Hey, I’m workshoppin’ some metaphors here!
Alright, we’ve entertained this monstrosity enough - time to wrap things up.
Rear Exterior
You know, McMansion Hell has been around for five years now, and has coined many terms - an art, ahoy matey, lawyer foyer, brass n’ glass, pringles can of shame - but I have to say, I hope fireplace nipples also sticks.
Anyway, that’s all for 1980 - join us next month for 1981.
Practically every animal that scientists have studied — insects and cephalopods, amphibians and reptiles, birds and mammals — can distinguish between different numbers of objects in a set or sounds in a sequence. They don’t just have a sense of “greater than” or “less than,” but an approximate sense of quantity: that two is distinct from three, that 15 is distinct from 20. This mental representation of set size, called numerosity, seems to be “a general ability,” and an ancient one, said Giorgio Vallortigara, a neuroscientist at the University of Trento in Italy.
Now, researchers are uncovering increasingly more complex numerical abilities in their animal subjects. Many species have displayed a capacity for abstraction that extends to performing simple arithmetic, while a select few have even demonstrated a grasp of the quantitative concept of “zero” — an idea so paradoxical that very young children sometimes struggle with it.
A few days after the media and public started to notice the works popping up in seaside towns around Britain, the artist has posted an Instagram video showing the graffiti pieces being installed, and what locals think of the new additions.
People have noted that when using giant internet-trained AIs like CLIP+VQGAN to generate images, you get much nicer-looking images if you include an artist byline.
Here's "Internet Infrastructure"
And here's "Internet Infrastructure by James Gurney."
James Gurney, a painter known for works like the Dinotopia books, is particularly prolific online, and people have noticed that adding James Gurney as a byline to an image tends to produce great results. Ryan Moulton's collection of generated James Gurney paintings is astounding.
In general, when it comes to generating images, it pays to know your art history. Here's a thread by hannahjdotca that showcases some of the range that CLIP+VQGAN has picked up from online art.
I've found that an artist's period does make a difference - "Internet infrastructure by Vermeer" (1632-1675) is a bit low-tech and hard to parse.
Similarly, "Internet infrastructure by Edward Hopper" (1882-1967) isn't great.
But "Internet infrastructure in the style of Edward Hopper" is much improved (and includes the AT&T logo).
I don't know if asking for an anachronism tends to produce bad results, or if I just picked two artists who don't lend themselves well to buildings and wires.
But I also happened to experiment with made-up artists, and surprisingly they had distinct styles.
Here's "Internet Infrastructure by Carmine Nottyors".
She has never existed. She paints a mean tangle of wires.
Here's "Cathedral of the giants by Carmine Nottyors".
Her style is distinct from the also made-up Picov Andropov.
And also distinct from the most similarly-named real painter I could find, Toni Carmine Salerno. (Here's "Internet Architecture by Toni Carmine Salerno").
I wondered if it was CLIP+VQGAN's figurative shrug at an unrecognized artist name. But check this out: here's "Sheep in Pasture by Carmine Nottyors".
And here's the clearly-different "Sheep in Pasture by Anonymous" ("Internet Infrastructure by Anonymous" was dominated by white masks)
Ironically "Anonymous" is the most prominently-signed painting of all.
I'm intrigued that CLIP+VQGAN can generate unique painting styles for nonexistent artists. It's a dimension of control that I don't think I've seen people exploring yet.
I'm not sure what it means that "Cathedral of Giants by Janelle Shane" looks like Devil's Tower crossed with butts.
For more cool-looking paintings in the style of James Gurney (as well as some hilarious flops), become an AI Weirdness supporter to get bonus content! Or become a free subscriber to get new AI Weirdness posts in your inbox.
Bonnie Berkowitz and Artur Galocha go with the strip plot to show the distribution of age for different Olympic events. If it’s longevity you’re looking for, go for the equestrian, sailing, shooting, and archery events. There’s still time.
Noah Kalina has been taking a picture of himself every day since January 11, 2000. He posted time-lapse videos in 2007, 2012, and 2020. Last year was the 20th of the project.
Usually Kalina’s videos are a straight up time-lapse using every photo. But in this collaboration with Michael Notter, 7,777 Days shows a smoother passage of time. Notter used machine learning to align the face pictures, and then each frame shows a 60-day average, which focuses on an aging face instead of everything else in the background.
So this has actually been cited by academics as part of the major draw to online spaces is the fact that just existing in public is reacted to with hostility and punishment. Gretchen McCulloch discussed this is in her book Because Internet, citing research that shows teens and young adults want to be outside! We want to spend time in social places, it’s just that there aren’t any places to exist in public without being charged for it.
When I was homeless as a kid my little brother and I loved to go to the library. We would keep warm in there reading good books all day long. Until residents of the town complained about us “loitering” at the library each day. The library staff then told us we were no longer allowed to stay more than an hour at a time. Imagine seeing two homeless children spending their entire days quietly reading just to keep out of the cold and having a damn problem with it.
Even the fact that teens use all kinds of social networks at higher rates than twenty-somethings doesn’t necessarily mean that they prefer to hang out online. Studies consistently show that most teens would rather hang out with their friends in person. The reasons are telling: teens prefer offline interaction because it’s “more fun” and you “can understand what people mean better.” But suburban isolation, the hostility of malls and other public places to groups of loitering teenagers, and schedules packed with extracurriculars make these in-person hangouts difficult, so instead teens turn to whatever social site or app contains their friends (and not their parents). As danah boyd puts it, “Most teens aren’t addicted to social media; if anything, they’re addicted to each other.”
Just like the teens who whiled away hours in mall food courts or on landline telephones became adults who spent entirely reasonable amounts of time in malls and on phone calls, the amount of time that current teens spend on social media or their phones is not necessarily a harbinger of what they or we are all going to be doing in a decade. After all, adults have much better social options. They can go out, sans curfew, to bars, pubs, concerts, restaurants, clubs, and parties, or choose to stay in with friends, roommates, or romantic partners. Why, adults can even invite people over without parental permission and keep the bedroom door closed! (page 102-103)
I often heard parents complain that their children preferred computers to “real” people. Meanwhile, the teens I met repeatedly indicated that they would much rather get together with friends in person. A gap in perspective exists because teens and parents have different ideas of what sociality should look like. Whereas parents often highlighted the classroom, after-school activities, and prearranged in-home visits as opportunities for teens to gather with friends, teens were more interested in informal gatherings with broader groups of peers, free from adult surveillance. Many parents felt as though teens had plenty of social opportunities whereas the teens I met felt the opposite.
Today’s teenagers have less freedom to wander than any previous generation. Many middle-class teenagers once grew up with the option to “do whatever you please, but be home by dark.” While race, socioeconomic class, and urban and suburban localities shaped particular dynamics of childhood, walking or bicycling to school was ordinary, and gathering with friends in public or commercial places—parks, malls, diners, parking lots, and so on—was commonplace. Until fears about “latchkey kids” emerged in the 1980s, it was normal for children, tweens, and teenagers to be alone. It was also common for youth in their preteen and early teenage years to take care of younger siblings and to earn their own money through paper routes, babysitting, and odd jobs before they could find work in more formal settings. Sneaking out of the house at night was not sanctioned, but it wasn’t rare either. (page 85-86)
From wealthy suburbs to small towns, teenagers reported that parental fear, lack of transportation options, and heavily structured lives restricted their ability to meet and hang out with their friends face to face. Even in urban environments, where public transportation presumably affords more freedom, teens talked about how their parents often forbade them from riding subways and buses out of fear. At home, teens grappled with lurking parents. The formal activities teens described were often so highly structured that they allowed little room for casual sociality. And even when parents gave teens some freedom, they found that their friends’ mobility was stifled by their parents. While parental restrictions and pressures are often well intended, they obliterate unstructured time and unintentionally position teen sociality as abnormal. This prompts teens to desperately—and, in some cases, sneakily—seek it out. As a result, many teens turn to what they see as the least common denominator: asynchronous social media, texting, and other mediated interactions. (page 90)
Anyway, more people need to read It’s Complicated, danah boyd really takes young people and technology seriously and doesn’t patronize or sensationalize, and it was a huge influence on me in figuring out the tone for Because Internet so I want to make sure it gets credit!
Inspired by space-saving flat-packed furniture, scientists at Carnegie Mellon University have developed a technique for making pasta shapes that start out flat when dry but “morph” into their final 3D shapes when cooked. The secret is stamping different groove patterns into the pasta dough.
The solution: something Wang, Yao, and their co-authors term “groove-based transient morphing.” They found that stamping flat pasta sheets with different groove patterns enabled them to control the final pasta shape after cooking. According to the authors, the grooves increase how long it takes to cook that part of the pasta. So those areas expand less than the smooth areas, giving rise to many different shapes.
The team found that the pasta reached its maximum bending angle after about 12 minutes and retained this angle for around 20 minutes before it began to bend back. The researchers were able to produce simple helical and cone shapes, as well as more complex saddles and twists (the latter achieved by introducing double-sided grooves).
Practically every animal that scientists have studied — insects and cephalopods, amphibians and reptiles, birds and mammals — can distinguish between different numbers of objects in a set or sounds in a sequence. They don’t just have a sense of “greater than” or “less than,” but an approximate sense of quantity: that two is distinct from three, that 15 is distinct from 20. This mental representation of set size, called numerosity, seems to be “a general ability,” and an ancient one, said Giorgio Vallortigara, a neuroscientist at the University of Trento in Italy.
It may look like a 
