Welcome to the latest edition of the Linkfest! Thank you for being a subscriber; if you’re enjoying it, spread the word – it’s a pay-what-you can signup here; the folks who can afford to contribute help keep it free for everyone else, too. Forward this email to anyone you want.
Let’s get to it! Here’s the best stuff I found online this week …
Art-generating AI firms – like Stable Diffusion – create their AI by scraping millions of images of artwork posted online, and using it to train a neural net. Artists are understandably pissed about this, particularly when the image-generators are used for “style mimicry”, where you prompt the AI to create a new image “in the style of” a known artist.
This got an academic team at the University of Chicago wondering: What if you could post your art online in a format that rendered it useless for training an AI – and useless for mimicking?
Lo and behold, they have a demo working. Called “Glaze”, it’s software that takes an image and alters some of its pixels in a way that confuses AI into thinking the picture is done in a different style (like, say, the abstract paint-blobs of Jackson Pollock) – so it can’t extract a clear signal for learning. Impressively, these perturbations can’t be seen by the naked eye; when a human looks at the image, they think it looks normal. But the AI gets mathematically confused.
The academic web site for Glaze is here, and that screengrab above is from their academic paper describing how it works.
The New York Times reports (free link, you don’t need to be a subscriber) on how the artist Karla Ortiz tried out Glaze on her work, successfully …
When Ortiz posted her Glazed work online, an image generator trained on those images wouldn’t be able to mimic her work. A prompt with her name would instead lead to images in some hybridised style of her works and Pollock’s.
“We’re taking our consent back,” Ortiz said. AI-generating tools, many of which charge users a fee to generate images, “have data that doesn’t belong to them,” she said. “That data is my artwork, that’s my life. It feels like my identity.”
In essence, this is an “adversarial” attack on neural-network AI. These attacks are becoming interestingly more common: You may have seen the ugly sweaters that, when you wear them, render you invisible to face-recognition technology? Or the turtle that was imperceptibly modified to make a visual AI think it was a rifle? Same thing there. This is one of the dangers of relying on visual AI for mission-critical things, by the way; researchers also found that placing a few stickers on stop signs rendered them invisible to self-driving car AI, yikes.
The academics note that Glaze is an imperfect solution and possibly/probably temporary; newer iterations of AI image-generators may figure out a way to get around this adversarial attack. But the general idea of adversarial attacks ain’t going away any time soon.
Google’s Street View has captured some curiously compelling pictures. Years ago I wrote for Wired about the artist Doug Rickard, who’d spend hours trolling along Street View scenes until he found arresting images, which he’d print up and put in a gallery: “These feelings of economic isolation and separation … they were like Hopper paintings.”
Now the web developer Neal Agarwal has created “Wonders of Street View”, a site where he’s curated 300 of the most interesting scenes he’s found. If you hit the “random” button it takes you to one of his picks, and if you like it you can click to see it on Google Maps itself.
Agarwal’s picks are pretty cool: I found a dolphin jumping,, a woman trudging along a highway, and – above – SpaceX’s Starship space craft. (Alas, I lost the direct link to that last one!)
“I think Street View represents our dream of wanting to travel the world from the comfort of our home,” Agarwal said. “I remember spending hours on Street View as a teenager exploring different countries and learning so much. I wanted the site to give a similar feeling to StumbleUpon, where it feels like you’re just stumbling on cool places in the world.”
… a photographic exploration of the beautiful design inside everyday electronics. Its stunning cross-section photography unlocks a hidden world full of elegance, subtle complexity, and wonder.
I want to get a copy of this book! I’ve long thought that the innards of electronics are some of the most strangely beautiful artificial objects. Some of the larger capacitors – large gel-colored tabs, or little chunky shiny canisters – almost look like candies, or perhaps jewels you’d see embedded on a bracelet.
In a review, veteran hardware-hacker Bunnie Huang marvels at how deftly the authors did photographic cross-sections of the electronics …
One of my favorite tricks of the authors is when they slice through optoelectronic devices: somehow, they manage to cut through multiple LEDs and leave them in an operable state, leading to stunning images such as a 7-segment LED still displaying the number “5” yet revealed in cross-section. I really appreciate the effort that went into mounting that part onto a beautifully fabricated and polished (perhaps varnished?) copper-clad circuit board, so that not only are you treated to the spectacle of the still-functional cross sectioned device, you have the reflection of the device rippling off of a handsomely brushed copper surface. Like I said: any engineering executed with soul and care is also art.
Huang also makes a lovely point about precisely why electronics components look beautiful:
Somehow, the laws of physics conspired with the evolution of human consciousness such that sound engineering solutions are also aesthetically appealing: from the ideal solder fillet, to the neat geometric arrangements of components on a circuit board, to the billowing clouds of standard cells laid down by the latest IC place-and-route tools, aesthetics both inspire and emerge from the construction of practical, everyday electronics.
The moon’s a big deal again, so there will be a ton of probes and astronauts arriving in the next decade. This poses a rather new challenge – of managing lunar traffic. You don’t want those projects slamming into one another, right? So the big space agencies realize they’ll need to build a GPS system on the moon.
But! Since GPS systems calculate position based on distance (how far you are from the nearest satellite) and time, the question becomes: What time is it on the moon?
This turns out be a difficult question to answer! As Elizabeth Gibney notes in Nature …
Defining lunar time is not simple. Although the definition of the second is the same everywhere, the general theory of relativity dictates that clocks tick slower in stronger gravitational fields. The Moon’s gravitational pull is weaker than Earth’s, meaning that, to an observer on Earth, a lunar clock would run faster than an Earth one. Gramling estimates that a lunar clock would gain about 56 microseconds over 24 hours. Compared with one on Earth, a clock’s speed would also subtly change depending on its position on the lunar surface, because of the Moon’s rotation, says Tavella. “This is a paradise for experts in relativity, because you have to take into account so many things,” she adds.
Go read that whole piece – it’s incredibly interesting. Satellite technology really makes relativity a pressing daily issue, even here on Earth: Because satellites are moving faster than we are, time goes slower for them, so Earth satellites have to make tiny constant time-corrections to keep them in sync with stations down here on the ground.
Back in 1985, drug smugglers dropped 40 containers of cocaine out a plane into the Tennessee wilderness. A black bear ate 75 pounds of cocaine and was later found extremely dead, its stomach “literally packed to the brim with cocaine”, as Wikipedia notes.
Now there’s a comedy-movie coming out depicting the cocaine-addled bear as going on a wild rampage. None of that happened, of course; the poor bear just died from a nutty overdose. Still, the flick looks like good messy fun.
To promote the movie, the directors have published an 8-bit-style game – a vague ripoff of Pac-Man, where you play as the bear, racing around a maze, devouring cocaine to speed yourself up as you chase several humans and messily devour them.
I’m kind of surprised more schlocky movies don’t make little throwaway promotional video-games like this? They used to do this a lot more back in the late 90s and early 00s, in the glory days of Flash. Ironically it’s probably now easier and cheaper than ever to crank out a quick branded video, what with the ease and low-cost of development kits like Unity or Phaser.
(via the Museum of Broken Relationships)
Zagreb, Croatia is home to the “Museum of Broken Relationships”, where people are encouraged to send in artifacts from relationships that have died. It was founded twenty years ago by Olinka Vistica and Drazen Grubisic, who, upon breaking up, couldn’t figure out who should get a knicknack that was meaningful to both of them. They decided to found a museum for such objects – an island of misfit love, as it were.
Alex Marshall wrote about it for the New York Times (that’s a “friend” link, so you can read it even if you’re not a subscriber):
The objects’ donors, many of whom send in the objects by mail, are kept anonymous, an attempt to encourage them to be truthful.
The mementos range from the mundane — such as a pair of white dress shoes, which came with a note reading, “She tried to impose her fashion sensibilities on me” — to the emotionally fraught. One exhibit is a parachute rig donated by a woman whose lover died in a skydiving accident. [snip] Charlotte Fuentes, a curator who looks after the collection and arranges temporary shows abroad, said that new objects arrived by mail each week.
Recently, she said, someone sent her a 37-year-old piece of wedding cake. She put it in a freezer.
“I’m amazed what people will do to try and keep love alive,” Fuentes said.
That toaster above? The donor called it “The Toaster of Vindication”, and submitted this text: “When I moved out, and across the country, I took the toaster. That’ll show you. How are you going to toast anything now?”
Interestingly, many objects are records of social upheavals – migrations and wars, which end not only lives but relationships:
“In the beginning, we were worried we’d just get items from summer flings, but the stories soon went deep,” Vistica said. “We’ve got items from the Second World War, about terrorism. Some of it’s heavy,” she added. “But life’s heavy.”
Sarah Ciston has published a book about machine learning – and the publishing platform is none other than … Google Docs.
This is pretty cool. I knew you could make any Google Doc public, which essentially means Google Docs is a content management system, and thus a weird sister of Wordpress or Tumblr. But this book of Ciston’s goes a step further – they’ve made it look beautiful, with the attention and care of a magazine or book layout. (The designers here are Vladan Joler and Olivia Solis.) I wasn’t aware you could coax Google Docs to allow such ambitious design!
(I’m mentally filing this alongside other curious experiments in repurposing Google’s online tools – like Marie Foulston’s “party in a Google sheet”, where you invite dozens of friends to edit a Google Sheet, live and simultaneously: Holding ad-hoc text-chats in adjacent cells, pasting in photos and memes, and generally hassling each other.)
The “shadow biosphere” theory. ⌛ A beer made out of air conditioner condensate. ⌛ Computer code that includes swear words tends to be higher quality, study finds. ⌛ Can satellites detect whales from space? ⌛ Colorado is aiming to become a hub of “agrivoltaics”. ⌛ “Mortal computation.” ⌛ Napoleon’s traveling library, the Kindle of the early 19th century. ⌛ The standing-waves explanation of Loch Ness Monster sightings. ⌛ A phenomenal essay explaining how ChatGPT works, by Stephen Wolfram. ⌛ UFO pendant lamp with a secret power outlet. ⌛ NASA uses AI to help design “alien”-looking parts for space probes. ⌛ Why modern pop songs have so many writing credits. ⌛ Smart punchbags. ⌛ Teardown of the electronics inside a 1980s D&D board game. ⌛ A machine-learning model to predict which parts of Sydney will gentrify. ⌛ Scientists create a black hole in a lab. ⌛ The moon smells like gunpowder. ⌛ Does thanking too many people in the credits mean a film is bad? ⌛ Type a message and Calligrapher.ai will deepfake it in handwriting. ⌛ “Quintessence”, a tool that shows you the most popular Mastodon posts among accounts you follow. ⌛ New research finds “a high correlation between documented Bigfoot encounters and dense populations of black bears”. ⌛ How whales’ fin-bumps inspired better windmill design. ⌛ The bike hunters. ⌛ “Jane Austen’s 8-Bit Adventure”