
You can’t spell AI without C-A-P-I-T-A-L-I-S-M
Around the Web
Issue No. 020
Not-so-news from your favourite AI shovel sellers, beaver bombing, how screen readers work, and the physics of riding a bike.
Welcome to Around the Web. It’s been a while.
In my absence the AI bros convinced the world that the singularity is nigh and a doom-machine impeding. «Pay us», they screamed in their finest snake oil salesmen voices, «or AI will destroy us all!» What a pitch! Welcome to the marketplace of doomsaying. But also: What a nonsense. Doom is mine, but I’ll share it.
Here we go.
This ain’t intelligence
In May, Sam Altman, CEO of OpenAI had a weasel-eyed appearance in front of the US Congress, asking pretty-please regulate AI; otherwise it’ll kill us all. Shortly after, he learned about the EU AI Act, which – of course – is actual regulation. Altman reacted by saying that this is the wrong regulation and OpenAI may stop operating in the EU if it isn’t changed.
You see, it’s a gold rush and as the shovel sellers are the only ones getting rich, it’s only fair that they decide how shovel selling in a gold rush will be regulated. They have the experience in the market after all, not those pesky politicians.
After all, they promise AI will destroy humanity if they – because they only want the best for humanity (which is a nice way to say their profits) – don’t build it responsibly. There is no evidence for such claims (except for the Terminator movies, which aren’t scientific literature, thinking about it). But this doesn’t stop ideological capture from taking over elite colleges.
These are the people who could actually pause it if they wanted to. They could unplug the data centres. They could whistleblow. These are some of the most powerful people when it comes to having the levers to actually change this, so it’s a bit like the president issuing a statement saying somebody needs to issue an executive order. It’s disingenuous.
Now, the FTC is investigating Open AI, too.
While OpenAI and Google try to make a secret out of their every move, data and development, Facebook decided to take a different route. They tend to be much more open with their models. For better or worse. Truth be told: For better and worse.
German tabloid Bild has cut jobs and is looking to integrate AI deeper into their workflows. As they just make shjt up anyway, a bit of hallucinating won’t make a difference.
Short interlude: Here’s ChatGPT failing at ASCII art.
The gay-detectors are at it again. A model that can detect homosexuality! Science! I’m so happy I’m not in Zurich right now. I especially like the quote at the end, where the honourable scientist is like «Of course we oversimplified, but we did it to showcase human diversity.» Slow clap.
On the bright side, we are seeing more and more journalists actually using their brain and looking deeper than press releases.
If you are a journalist, or know a journalist, or just want to be able to critically read what journalists write, the Columbia Journalism Review had an actual useful step-by-step guide on how to report better on artificial intelligence.
Intelligencer published a must-read piece on the human work that underpins every Machine Learning model, no matter if it trains itself or relies on annotated data. This work isn’t going anywhere and as the companies developing the models have understood capitalism quite well, it’s low-paid and outsourced.
But behind even the most impressive AI system are people — huge numbers of people labeling data to train it and clarifying data when it gets confused. Only the companies that can afford to buy this data can compete, and those that get it are highly motivated to keep it secret. The result is that, with few exceptions, little is known about the information shaping these systems’ behavior, and even less is known about the people doing the shaping.
Those doing the work have formed their first union. Here’s to many more to come.
Andrew Deck, in Rest of World, looked at outsourced and contract workers in the majority world. They are the ones who’ll will take the brunt of AI’s impact and are adapting to generative AI already. The takeaway here is that AI won’t make humans redundant, but workflows will change.
Black artists are investigating the datasets and outputs of generative models and the way those models are (not) able to reproduce authentic images of Blacks. Those models distort their faces, or lighten their skin. Meanwhile, the terms of service hinder research, as e.g., the generation of images depicting slave ships is blocked. For good reason, as we all know that certain corners of the internet will do with it. At the same time, this block makes it impossible to investigate some pieces of human history.
Those datasets, also include tons of private data – to the surprise of no-one. An analysis by the German public broadcaster Bayerischer Rundfunk found thousands of files with intact EXIF Metadata in the LAION-5B dataset. EXIF data can contain names or geolocations, enough to deanonymise people on photos in the dataset. Deleting such metadata isn’t hard, and not doing it a colossal oversight. As LAION is public, there likely is a multitude of models trained on the data already.
Google’s new «Search Generative Experience» might break the internet, argues Avram Piltch in Tom’s Hardware. Google is set to plagiarise content it finds, deprioritising actual search results further and further. The next time a Googler explains how much they care about the open web, you are allowed to laugh them out of town.
Maybe it’s time to lock off GooBingAI from our websites? «We can rescind our invitation to Google», says Jeremy Keith. And actually, we should focus on building corners of unadulterated humanity, hidden away from corporate crawlers.
Mozilla, nowadays unfortunately always one to ride a wave, even if it’s only one to embarrassment, tried to use some Chatbot model on the Mozilla Developer Network. It didn’t end well.
Okay, here’s something pretty impressive. Redditor nhciao posted a series of graphics where they used Stable Diffusion to embed working QR codes in anime images. Mindboggling stuff.

EOL of Humanity
While Uruguay is suffering its worst drought in 74 years, with the government even mixing saltwater into the drinking supply, Google plans to build a data center in the country using ever more fresh water. «At Google, sustainability is at the core of everything we do», says Google, while Uruguayans say «This is not drought, it’s pillage».
Some months ago, the EU made headlines, when they planned to ban most hazardous chemicals. At the tides of time and some millions of lobbying money: EU to drop ban of hazardous chemicals after industry pressure. The mimimimi of capitalists is the most pathetic sound of our century.
Meanwhile, in the Nature is Healing Department: A renegade sea otter is terrorizing California surfers: «It’s a little scary». Also, the orcas are still out to get us, and I’m in the cheering section.

If I’d ask you to explain beaver bombing to me, what would you come up with?
If it’s releasing beavers into habitats to reverse their extinction that is indeed the correct answer. If you thought about something else, I did too. The whole process is illegal and some say may harm the environment. Beaver bombers don’t care.
“I had all the authorizations I needed,” Rubbers said. “Which, in my mind, meant no authorization.”
“He bombs quite a bit,” Schwab admitted about Rubbers. “He wanted to do something for nature.”
Paris, at one point a city engulfed in car emissions, piece by piece manages to become a greener and healthier city.

Social Mediargh
The biggest story over the past few months was probably Reddit’s update to its API policy. In short, you’ll now have to pay good money to access Reddit’s content for, say, an app you are building. Which sounds nice, but is a threat to a whole host of 3rd party providers who built their apps around the free API available until now. It kinda sounds reasonable, except there has been as little as no warning and Reddit lives of free work of others.
In protest, Redditors made their subreddits private, the mod team of the Ask me Anything subreddit announced they will cut down their volunteer work drastically. Reddit, so far, remained sturdy and said they won’t budge.
Because everything is going great, Facebook finally decided to launch their long «awaited» real-time messaging thingy called Threads. Names are dead, let’s just put something generic on our product. Old Elon wasn’t amused and threatened to sue Facebook. Which probably left some litigators in Facebook’s House of Litigators pretty amused.
Of course, as it’s technology, and fuxk you, disabled people, Facebook «missed» some crucial accessibility features such as alt text for images. It’s well done all around, you see.
In Kenya, Facebook is busy busting unions, fortunately the courts have none of it.
As Facebook launched Threads, Zuckerberg launched Hobbyist Zuck, the brand-new version of himself, trying to get rid of his bland cyborg alter ego.
Now, you may be tempted to feel something like sympathy for Zuckerberg. Human after all? Don’t be fooled. Cat Valente was kind enough to write one of those cathartic pieces of contempt that are refreshing, like a bath in a mountain lake.
Well, it’s pretty fucking weird how the launch of Threads, which is ostensibly, you know, a company and a profit-generating service, almost immediately did a sickening costume reveal and became Mark fucking Zuckerberg’s Redemption/Woobiefication tour, and only like four non-Nazi people and one of their alt accounts are pushing back on that because everyone rushed to join this thing with a smile on their lips and a song in their heart a big anime heart-eyes for the guy we all knew was Noonian Soong’s first janky and obviously evil Build-a-Bloke workshop project three weeks ago.
Seriously, have we all lost our entire screaming minds?
I’d like to quote it in full, but I trust in you, dear reader, to read it anyway. Go on, I’ll wait.
Speaking of «friends», is anyone left in the Metaverse?
What are you looking at?
France is set to allow police to spy on suspects through remote phone access. But there is nothing to worry about, as it’ll only affect some cases and only after a judge allowed it, and politicians never in the history of ever tried to change this after the fact. Except all the time.
Speaking of the history of ever: Chat control, the zombie hunting encryption for the children, is refusing to die. The EU member states are now at the brain worm state where they say they would rather not weaken encryption but read encrypted content. How? Who knows! Sadly, we might find out. Of course, all of this is dogshjt:
The proposed regulation would also set a global precedent for filtering the Internet, controlling who can access it, and taking away some of the few tools available for people to protect their right to a private life in the digital space. This will have a chilling effect on society and is likely to negatively affect democracies across the globe.
Maybe the EU should have a ’lil look at the spyware vendors. LetMeSpy (an app you hopefully have never heard about) was hacked and got all their data stolen and published.
The German Ministries of the Interior decided not to use Palantir’s software, as it’s too bad. Bavaria, meanwhile, was like «We got a password, can share the account.» Peter Thiel will not be amused if you do the Netflix, I guess.
Hard times to be a surveillance vendor. Perhaps sell your assets! That’s what NSO Group tries to do. The Guardian reported that a producer of Adam Sandler movies and an heir to the Wrigley family (of chewing gum fame) are interested in its assets. Of course, the NSO Group and its Pegasus spyware can’t be used in the USA, and as such the Nation Security Council responded with «do whatever but do not think we’ll be impressed» (not a direct quote).
Adam Sandler. Chewing Gum. Pegasus. At one point, I might try not to feel like I’m in a fever dream and everything is just plain stupid when writing this newsletter. I have close to zero hope this will happen.
Loose ends in a list of links
When talking about the EU border, the first organisation that comes to mind is likely Frontex. In its shadow there is the International Centre for Migration Policy Development, a non-EU but EU-funded organisation, helping coast guards to keep refugees out of Europe. Coda explained how an EU-funded agency is working to keep migrants from reaching Europe.
I tend not to understand low-level technical content, and as such, I’m always super happy if there is deeply technical content I can actually understand. Neill Hadder published a series of three articles explaining how screen readers actually work. Here’s part 1, Swinging Through the Accessibility Tree Like a Ring-Tailed Lemur. But make sure to not miss Part 2, and Part 3.
I promised myself that I’ll finish this epic about Ticketmaster and its dark history (spooky but true) at one point. But, honestly, I started in January and progress is slow. So, I might not. If you are into music and the way live music has been monopolised, it’s a worthwhile read.
The Serotonin Signal succinctly explains the current state of science around the impact of Serotonin on depression. The gist: While Serotonin levels are not linked directly to depression, there are downstream effects that Serotonin does have which are linked to depression. The human brain is a pretty darn complex and wonderful thing.
And, finally: The physics of riding a bicycle. It’s the latest piece by Bartosz Ciechanowski (regular readers might remember his explanation of a mechanical watch), and it’s great as ever.
That’s it. Writing more than two kind-of-coherent sentences at once felt pretty good, albeit exhausting. While I can’t make any promises to return to a steady publishing rhythm as long as my lovable but flawed brain recovers, I’ll try my very best to at least have a publishing something.
In the meantime: Stay sane, hug your friends, and ride your bicycles fast and slow, far and near.