[links] Barbie used too much R343 Neon Pink / AI-Doppelgänger-Phobia / Isopod Ramen
Plus: Black Mirror S06 / Daniel Dennett on counterfeit people / ChatGPT draws an ASCII-bunny / Robot chef makes salad / Turing Trains / Solarpowered billion year Lego-clock
If you read this Newsletter for a while, you know i’m not an AI-doomer and i think those “existential risks“ and warnings of “extinction“ are an overblown part of the hype machine. But i also think that there are real psychological AI-risks due to it’s mimetic nature, where AI-models more often than not resemble real or imaginary people. Now there’s a new paper on Risks of AI Clones to Selfhood and Relationships: Doppelganger-phobia, Identity Fragmentation, and Living Memories, and it’s exactly about the stuff i’m getting at here:
“We found that (1. doppelganger-phobia) the abusive potential of AI clones to exploit and displace the identity of an individual elicits negative emotional reactions; (2. identity fragmentation) creating replicas of a living individual threatens their cohesive self-perception and unique individuality; and (3. living memories) interacting with a clone of someone with whom the user has an existing relationship poses risks of misrepresenting the individual or developing over-attachment to the clone.“Related: In a new paper on the “effect of social robots’ anthropomorphization on empathy towards human beings“ researchers found that we naturally prescribe a mind to a mimetic AI-model, develop a theory of mind for chatbots and that this synthetic theory of mind reduces empathy for human beings.
Also related: Computational model mimics humans' ability to predict emotions.
A few month ago i wrote about how mimetic AI-models which simulate the use of language are a door opener for a synthetic theory of mind and how this can lead to new forms of self-radicalization. Predicting emotions in other human beings is pretty much a mainstay of your theory of mind, and if machines can do that, nobody will be able to not anthropomorphize them, which, as in 2, seems to reduce our empathy towards other people.
This is why especially this more subtle “make the machine more human like“-stuff is concerning me a great deal. Paperclip Maximizers might be great at parties, but humanlike mimetic AI actually gets to your head.Still related: In the latest Machine Learning Streettalk, Dr. Tim Scarfe speaks with renowned philosopher Daniel Dennett about the potential dangers of AI and the concept of "Counterfeit People." Dennett raises concerns about AI being used to create artificial colleagues, and argues that preventing counterfeit AI individuals is crucial for societal trust and security. They delve into Dennett's "Two Black Boxes" thought experiment, the Chinese Room Argument by John Searle, and discuss the implications of AI in terms of reversibility, reontologisation, and realism.
Still related: Marisa Tschopp at TEDxBoston: Friends with AI? It's complicated: "if people don't actually love their voice assistant but also don't see ‘her’ simply as a tool, how do they perceive their relationship with conversational AI? And what can we learn from better understanding human-AI relationships?"
All of these first five links point at one thing: Conversational AI will influence our behaviour through psychological mechanisms we can not escape, which exposes our very social fabric, psychological makeup and personal intimacy to algorithmic processing and infosec threats. I do think that this is the most imminent AI-risk we should talk about more.When italian authorities banned ChatGPT for privacy violations, “the output of Italian developers decreased by around 50% in the first two business days after the ban and recovered after that“. That’s quite a number, isn’t it.
Synthetic engagement numbers incoming: AI learns to use graphic user interfaces from screenshots.
This tweet featuring a synthetic Bill Gates talking to AI-Socrates is making the rounds for a while and it got dunked on pretty hard already for asking seriously if this is the future of podcasts. My short take: I linked to one of these synthetic podcasts a while ago and never listened to it again.
My longer take: Ideosyncrasy is the most valuable asset. With mimetic AI becoming better by the minute and podcasters, for instance, training genAI on their voices and talking style, people will listen to whoever can distinguish human recordings from other cues than quality of voice. The only reason people actually listen to AI-Bill Gates vs AI-Socrates because it's a gimmick and a cool show off of those shiny new tech abilities. Nobody listens to this because a robot said anything interesting.
AI-mimicable talking style does not include ideosyncratic ideas that express themselves in strange metaphors, weird topics, short and fast jokes related to the topic on a esoteric level, fun neologisms or just plain stupid crap --hallucinations be damned--, and so forth. Not sure if AI can ever solve that. Maybe it can, but that will require much more than training an AI-voice-clone.
Podcasts are interesting because people are interesting and they show interesting ideosyncratic ideas and unexpected choices. AI fails at both, which is why this is not the future of podcasts for a very long time, no matter how good the quality of voice will be."The Frost is a 12-minute movie in which every shot is generated by an image-making AI". And it does not look very interesting, it’s just illustrations generated with image synthesis animated with AfterFX. I want my AI-cinema to be true generative video, not sophisticated comicbook-trailers.
Some rad ASCII-art by Adel Faure.
Related: After visiting above link, i asked ChatGPT to draw a pixel bunny with some ASCII-characters. It happily complied, and the bunny looks as expected. I also tried to make it draw a map for a roguelike RPG game, and the results are equally… debatable.
3D-Printed Eigenfaces — Eigenfaces are statistical representations of basic face features for face recognition in AI-systems, and i want this 3D-printed set framed and hanging on my wall.
StyleDrop: Text-To-Image Generation in Any Style — New technique by Google that combines style transfer and image synthesis. Lots of examples in this Twitter-thread.
The Image Journal has a whole series of papers on image synthesis from the perspective of media theory, featuring some people i’ve mentioned and written about here (hello Roland, hello Hannes, hello Eryk!)
The Ukraine war sure is a war of newschool deepfake propaganda techniques: Russia Says 'Fake' Putin Address Declaring Martial Law Was a 'Hack'.
If you’re wondering how on earth AI could undermine democracy — it’s just a chatbot after all, amirite? — here’s a plausible scenario how this can work.
A robot chef watched some humans prepare food and made some salad.
Isopod Ramen for a thousand yuan per bowl. No i'm kidding, this is just me before, during and after enjoying a nice session with some bathing salts in the spa.
RIP Rargb, one of the oldest warez sites and this is sad because they were one of the rare release groups caring about dubs of foreign language movies. (I hate subtitles.)
Turing Trains is a website about the “design and construction of train track layouts capable of binary computation“. I need to up my model train game.
Careless Whisper's lick on 100 instruments (but without saxophone)
If you know clicker games, don’t play Array Game. Don’t click on this, under no circumstances. It’s the most addictive thing since Swarm Simulator. Don’t click on this one either.
“Brick Technology built a solar-powered Lego clock that will keep time for a billion years. It's got various displays in the style of an astronomical clock so you can keep track of seconds, hours, months, centuries, and even galactical years (the amount of time the Sun takes to orbit the center of the galaxy). The clock is powered by solar energy, and the solar cell is connected to the clock so that it tilts throughout the day to keep facing the sun.“
Hidden in an interview about the architectural design of the Barbie house from the movie, there’s this gem of an info-nugget: Barbie's Movie Had Enough Pink to Cause a Worldwide Shortage. Apparently, the production crew used so much of the Roscolux R343 Neon Pink that the makers ran out of paint: “The world ran out of pink”.
They should’ve asked Stuart Semple who made the pinkest Pink some years ago, an “ultra-bright paint“ with a “strong fluorescent effect“.
“John Carpenter stated there ‘may’ be a sequel to The Thing in development but the director is ‘sworn to secrecy’.“ — Sounds like they are going the Halloween Kills/Ends route for The Thing, too, with Carpenter somewhat involved but neither directing nor writing. Like every horrorhead, i hold The Thing close to my heart, and while i thought the last two Halloween-films were fine, i just hope they don’t pull something too ridiculous with this one.
Robert Eggers’ ‘Nosferatu’, with Willem Dafoe and Bill Skarsgård, wraps Prague shoot — When i was twelve years old or so i got my first library membership and i watched plenty from their movie section, including a lot of silent movies ranging from The Cabinet of Caligari to The Golem to, well, Nosferatu. I think i’ve seen the bald Dracula-ripoff more often than the original, and Max Schreck just is the most frightening vampire, ever, fullstop. So, i’m quite anxious about this one, but: I love every movie Robert Eggers has done in his short carreer. All of them were fresh takes, original and visually outstanding, and with Black Philip the goat he showed us one of the most interesting depictions of the devil, ever. If anyone can pull a Nosferatu-remake whose name is not Werner Herzog, then it’s Robert Eggers.
I’m a Virgo by Boots Riley is a new series about “a 13-foot-tall 19-year-old Black man raised by his Aunt Lafrancine and Uncle Martisse in Oakland, California.“ I have a knack for movies about weirdly scaled humans, and i think that “The incredible shrinking man“ is one of the best scifi-flicks ever. This series with it’s absurdist take on the giant-people thing is right up my alley. Here’s the trailer:
The new Black Mirror-trailer, also added to my short post on the series.
