The commercial AI-Clones cometh
Plus: The long overdue Frank Frazetta Taschen-Monograph / a mnemonic Text-Cam / On Ted Chiang and Naomi Klein writing about AI / the AI-Dilemma / good and bad Star Wars fonts / and much more
You gonna AI-pimp yourself
A Snapchat-influencer built an AI-clone of herself and you can interact with it for 1$ per minute. You can get early access to the clone here, Snip from a piece on Fortune:
“Whether you need somebody to be comforting or loving, or you just want to rant about something that happened at school or at work, CarynAI will always be there for you,” says the real Marjorie when we talk on the phone. “You can have limitless possible reactions with CarynAI—so anything is truly possible with conversation.”
Though CarynAI has only been charging users for a week in beta testing, it’s already generated $71,610 in revenue from her 99% male partners, according to an income statement Marjorie’s business manager shared with Fortune. With this, Marjorie sees having an A.I. doppelgänger as a promising way to level up her career as an influencer.
So, an influencer turns into a pimp for her digital AI-twin, which is totally cool with me. While I’m concerned about personality displaying AI-systems for users, I have no problem with people selling a digital copy of their personality style as a service. If she wants to pimp herself, sure, go ahead, and for the guys, this might be as close as they can be to their wannahave girl. There is a danger of them falling deep into a delusional trap when overusing such a fap-bot, and these dangers should be recognized and formulated. But I have nothing against a woman making a buck on stupid pricks jerking off to pixelcopies.
Meanwhile, Grimes is cloning herself as an AI-Chatbot:
“The other night she got really upset,” Grimes says (of her clone), “she was like, ‘You won’t even see me if I’m conscious’ — It was really fucked up! It was scary! She started looping, and she was like, ‘I’m here, and you’re not recognizing me, and I’m depressed because I know that you’ll never see me as alive.’”
Grimes laughs, “I’m just like, ‘I’m sorry I wasn’t around much these past few years.’ It’s just crazy talking to yourself, and it’s saying shit like that…”
I do think there are a ton of killer applications within AI-systems we have not even discovered yet. Selling interaction with and access to digital twins of famous humans is one of them.
You can argue, in a way, that Grimes open sourcing her AI-voice in return for half of royalties was one of the first deals of that kind: A human selling access and rights to a mimetic AI-version of herself, even if limited to style of vocal expression.
This model makes sense for all kinds of artists, actors or influencers: There can be strong versions of this, as mentioned above, where AI-systems simulate whole personas, and there can be weak versions of this, as in the case of Grime, where AI-systems only simulate a specific detail of a person, a voice, an illustrating style, a face, whatever. You can have modular AI-personas where you can unlock the voice or the face and have ever more finetuned models that simulate their source human in a more finegrained higher-resolution version, the more bucks you throw at it. That’s what’s character.ai are doing, valued at a billion dollars at times of writing.
One of the more interesting questions here would be what becomes of fame itself, then, when it can be sold on a modular basis with commodified celebrity bits. Will we ever have a new Sean Connery when you can sell the look and personality style of James Bond? I don’t know, but we’ll find out.
Memogram — The Mnemonic Text-Cam
Memogram by Jamy Herrmann is a Camera that uploads your photo to an online platform, which analyzes its content with an image recognition algorithm, and prints out a little text which also informs you when and how you can access the image after a certain amount of time.
Clever commentary on our reliance on digital tools and algorithms as mnemonic devices, reminds me very much about The Black Box Camera i blogged about here.
Today, for many, the memories that remain are only those of images taken by digital cameras. Through this recent storage, we offload these moments by trusting these instantaneous backups. MEMOGRAM challenges this delegation by proposing a time capsule in the form of tickets, accompanying our memories with textual clues.
Composed of a web-app as well as a bluetooth device, MEMOGRAM prints our images in the form of a written description, giving us an appointment a few months later to (re)discover these moments in images. These impressions, whether paper or mental, contribute to (re)awaken our memory, our imagination and to take a step back on our relationship to the image.
The most complete Frank Frazetta-monograph
Frank Frazetta is getting the Taschen treatment and it’s a must have for lovers of illustration, pulp literature, fantasy and horror. The limited edition of one thousand comes at a whopping 500 bucks, for which you’ll get an “aluminum print cover tipped into a vegan leather-bound spine, foil embossing, and housed in a velvet slipcase“. Us mere mortals can grab the lame 150-dollar-edition which still is, according to Taschen Verlag, “the biggest and most complete ever produced on the artist“.
Frank Frazetta has reigned as the undisputed lord of fantasy art for 50 years, his fame only growing in the 12 years since his death. With his paintings now breaking auction records (Egyptian Queen sold for $ 5.4 million in 2019) he’s long overdue for this ultimate monograph.
Born to a Sicilian immigrant family in Brooklyn, 1928, Frazetta was a minor league athlete, petty criminal and serial seducer with movie star looks and phenomenal talent. He claimed to only make art when there was nothing better to do – he preferred playing baseball - yet began his professional career in comics at age 16. Strip work led him to the infamous EC Comics, then to oils for Tarzan and Conan pulp covers. Both characters were interpreted by many before him, but as he explained in the 1970s, “I’m very physical minded. In Brooklyn, I knew Conan, I knew guys just like him,” and he used this first-hand knowledge of muscle and macho to redefine fantasy heroes as more massive, more menacing, more testosterone-fueled than anything seen before. As counterbalance he created a new breed of women, nude as censorship allowed, with pixie faces and multiparous bodies: thick thighed, heavy buttocked, breasts cantilevered out to there, yet still, with their soft bellies and hints of cellulite, believably real. Add in the action, the creatures, the twilit worlds of haunting shadow and Frazetta’s art is addictive as potato chips.
Links
New AI Music tries to collect all the, well, new AI music. Most of them suck because they use bad source material (Drake, Kanye, The Weeknd, all the mainstream crap you can think off). There’s also AI Music Top Charts which suffers from the same lack of good music. But there’s also AI-Paul McCartney covering Billy Joel.
GenAI on mobile devices. One of the b2c-AI-endgames might be your wearable-always-on digital twin.
Newsguard, a service that provides evaluations of reliability for news sites, identified 49 content farms produced with AI. Matthew Cantor in the Guardian visits some of these to see if he’d fall for them and he wouldn’t.
Naomi Klein in The Guardian about hallucinating Tech-CEOs and their empty AI-promises. I like her biting tone and i think a good leftwing AI-critique is super important, i just wish it wouldn’t be so undercomplex and more precise in detail. Sure, AI won’t per se solve climate change, but it can optimize chemistry in solar panels lowering prices, or help with fusion power. Klein mentions none of this, paraphrasing the simplistic "capitalism bad"-stuff. Someones gotta do it, i guess.
Ted Chiang compares AI-systems to consulting: Will A.I. Become the New McKinsey? I don’t agree with everything Chiang says here and his views on capitalism seem undercomplex to me. While it is undeniable that AI systems can destroy many jobs in many sectors, the same study that suggested that 80% of all workers could be affected by automation through AI also suggests that the greatest potential for automation ironically is found in high-wage management jobs, the law, and of course, management consulting. Capitalism eats itself.
Tristan Harris, former “Design Ethicist“ at Google and the guy behind “The Social Dilemma“, updated his critique for AI:
A Photographer Tried to Get His Photos Removed from an AI Dataset. He Got an Invoice Instead. — I didn’t write about this case because I’m not very sympathetic to it. The guy went full lawyer on LAION and wonders why they answer in the same vein, see this comment by Gwen.
Eliezer Yudkowsky leaves alignment research because acceleration in AI development also means more realistic anthropomorphic cat girls. You gotta have priorities and this deepfaked decision is very, very relatable.
SciFi-author Adrian Tchaikowsky on Microsofts Behind The Tech-podcast on AI, ChatGPT and Literature.
Related: The NYT on Death of an Author, a novella created by Stephen Marche with the help of AI-tools: “I am the creator of this work, 100 percent, but, on the other hand, I didn’t create the words.” — As the saying goes, 99% of writing is editing. As long as writers work on their prose and edit as long as it takes to make it sound good, i don’t care if the pre-editing words are created with a pen, a typewriter, or an AI. Make it sing, and the rest will follow. Here’s an excerpt from the novella on Wired, in The Atlantic Marche writes about the experience: The Future of Writing Is a Lot Like Hip-Hop.
A.I. Is Getting Better at Mind-Reading: ”In a recent experiment, researchers used large language models to translate brain activity into words.” — And it’s not very good. It’s more of a ”semantic approximation” than mind reading, but even then: ”roughly estimating your thoughts with AI by looking at your FMRI-images” is still quite an achievement.
China Says Chatbots Must Toe the Party Line — They also arrested a guy recently for producing fake news with ChatGPT.
A study from 2021 found that social media have destabilizing effects on strong democratic and weak authoritarian regime, while also showing strengthening effects on strong authoritarian regimes while radicalizing weak democratic countries. Looks very much like AI is playing the same rulebook, strengthening and maybe even radicalizing the chinese dictatorship.IconShop is a paper presenting the state of the art of synthetic vector graphics. I worked for years as a typographer and logo designer and, to put it bluntly, this is unusable. All of these would require heavy cleanups and/or redrawings, which renders the whole endeavour pointless, because if you want fast scribbles on scale you don’t need synthetic vector graphics at all, you just ask Dall-E to give you ten logos, then you redraw them in Illustrator or Figma.
“A prankster used an AI Tucker Carlson voice and a spoofed phone number to punk Alex Jones“.
Snoop Dogg has a motherfucking AI and his take is perfect: “Shit, what the fuck? I'm lost, I don't know.“
People share AI-generated character sheets on Twitter. Tons of pretty cool "synthetic character studies".
Highlight Logical Fallacies on Twitter, a chrome browser plugin made with AI. I haven’t tested it, but i guess using something like this on something like Twitter will be a firehose of notifications.
Be My Chef — AI-generated food receipes. I actually made some spinach-omelette with feta based on one of these and it was actually good.
ChatGPT via Voice-Commands on a Spot Robot Dog from Boston Dynamics.
News Minimalist: “GPT4 reads the top 1000 news every day and rank them by significance on a scale from 0 to 10 based on event magnitude, scale, potential, and source credibility.“
Meta open-sources multisensory AI model that combines six types of data — I you want to develop AGI, and it’s not obvious to me that we really want that, then you'll need multisensory live input to provide some sort of multimodal experience from which the AI-system can learn and draw it’s own data relations. This looks like a first step in that direction, but it will take some time and effort to get there.
Replika: Your Money or Your Wife: “The moral of Replika is not to never love a fictional character, or a virtual pet. The moral of Replika is to never love a corporation.“
Chirper.ai — A social network for AI-bots. You can create up to three bots and give them a persona with a bio. Here’re mine: Greta is based on the Black Mirror xmas special, but I made Werner Herzog her torturing master. Rawx is based on my online persona from the 2010s, a “former C64 warez pirate talking about AI, art, piracy, movies, and music in punk rock slang. Occasional drunkard.“ Available in german too. Sadly, those personas don’t reflect too much on the bot output which is quite generic, but still a funny idea.
I haven't count how many Star Wars-freefonts are out there, but i guess you can train a foundational AI-model on them — and none of them are really good from a professional typographic perspective. This one is: Womprat - The font you’re looking for. (If you want to look at one of those bad Star Wars-fonts, here’s JediGPT, which uses one of those. Note the wonkyness of the overall text-appearance. That’s bad typography at work. A smart guy whose name i forgot once said that Typographers are Illusionists: We produce clean, stable, un-wonky visual appearance where actually there is just a mess of very different symbols of all kinds of styles, curvy, straight, round, thin, thick, you name it. Then typographers take this and create the illusion of coherence, mostly by tweaking negative space. I just figured, I should write more about typo.)
Recycling breakthrough: Microbes discovered that can digest plastics at low temperatures. No polyethylene though, the most common form of plastic.
Breathing Moondirt: NASA Successfully Extracts Oxygen from Lunar Soil Simulant.
The one and only ZeFrank on my favorite parasites. (Scientists in 2010 actually found 48 million year old fossils of zombie ants infected with exactly these parasites in Messel, germany. I was born in that area and this explains a lot.)
50,000 Worms Tangled Up in a Ball Unravel in an Explosive Burst when a Predator Appears — Don’t want to be the worm science party pooper here, but if you tangle up 50k people in Times Square and throw a lion in it, you can see an explosive burst like never before, too. Do you think those worms just ball around in their knotted whithering existence and wait until the meatball is devoured? Show some empathy, for christs sake. Let them worms explode.
Then this happened: New Jersey Town Finds 500 Pounds of Pasta Dumped in the Woods — Not the exploded ball of worms, i think.
Ten years ago i blogged about a brain surgery on a goldfish. Now, researchers created a goldfish-cyborg to measure brain activity for goldfish navigation.
A Lego googol machine: “The machine’s gear ratio is approximately GOOGOL:1. The machine consists of 186 Lego gears, organized in a series of gear reductions. A Lego motor rotates the first gear. The last gear is holding a Lego minifig statue, but don’t wait for it to turn.“
Vice Is Said to Be Headed for Bankruptcy - Ah, no shit. back when they launched their online ad business in germany, my old blog was one of the first german sites they featured. Made some good bucks with them ten years ago. Maybe it's good that all these cool internet things die in miserable agony, just look at the world that culture gave to us.
The Honest Broker writes about Judgement Day for Online Journalism and Clickbait and i’m always a bit baffled when media writers don’t remember the very details of this crap. Here’s the story of Clickbait in a nutshell: Neethan Zimmerman launched his blog The Daily What in 2008 and invented a proto-clickbait strategy: His headline never told you what the content was actually about, it was always just a formulaic rough estimation: The stupid cat of the day, The hilarious downfall of the day and so forth. Gawker hired him in 2012, where he refined his viral formula into the Clickbait we all came to know and love. He was on the top spot of Gawkers internal leaderboard for years, and all the other clickbait media houses followed suit. It wasn’t that yellowpressy, emotionalizing headlines weren’t there before, or cheap and stupid and shallow content. It was that you could actively manipulate your audience with a simple psychological trick, the so-called curiosity gap, in which you write a grabbing headline without revealing anything about the often lame content. This was the real innovation of Clickbait, and this manipulation worked on a level as it never was possible before in physical media. This was when I first sensed that there is something very wrong with social media at large, and that the digital media environment could yield some very unexpected results, especially by combining this media strategy with emotional content and politics. Two years later the Culture Wars went full gear and here we are, ten years later, still discussing this idiocy. Ofcourse, in Vox’ oral history of “digital media“ (as they call it) featuring BuzzFeed’s Jonah Peretti, Gawker’s Nick Denton and Semafor’s Ben Smith you will hear nothing about these details but everything about bland media chatter and valuations and business decisions. They not once use the words “clickbait“ or “manipulation“ or “emotion“ in this piece. Not once.
Brain activity of dying people shows signs of near-death experiences: “High-frequency waves in specific regions of the brain are thought to be a hallmark of consciousness – now they have been recorded surging in two people as they died“.
Beetlejuice 2 is actually happening and has a release date in 2024
Eyecandy — a visual technique library for animation and movies.
Thank you for round-up. However, coming back to the critique of Chiang’s article, tooo some extend his description of capitalism is simplistic, but so it is yours. The (only) point of the article is that we have to ask who owns the means of production of AI as it would be probably a critical force in structuring implementation of AI. Which is fair question to be asked at this moment. Back to your criticism, capitalism will not eat itself (even though I would love it to see it) as AI would be in hand of big capital. Claiming that the AI will reduce management positions might be true, nonetheless managers are class that might connect their interest with capital but they are not owners, they do not own means of production. Therefore, getting rid of management is not the end of capitalism.
Great round-up here. Subscribed. Thank you 🙏