[Links] David Cronenbergs "Galaxy of Fear" (1985)
Plus: When machines change art / Report on automated influence operations / Negligible impact of Putins Troll Army / A "This is fine"-sweater / and much, much more
An invisibility cloak for computer vision: “This stylish pullover is a great way to stay warm this winter, whether in the office or on-the-go. It features a stay-dry microfleece lining, a modern fit, and adversarial patterns the evade most common object detectors. In this demonstration, the YOLOv2 detector is evaded using a pattern trained on the COCO dataset with a carefully constructed objective.“
Aaron Hertzmann has a long, very good blog post about AI-art and its ramifications from ethics to aesthetics: When machines change art.
New report on the use of AI for “automated influence operations”, in which the authors suggest to ‘nuke the internet’ with ‘radioactive data’, to contaminate the training data to make it detectable. I don’t buy into the flood, and, funny enough, two days before this report was published, researchers in a new paper found “no evidence of a meaningful relationship between exposure to the Russian foreign influence campaign and changes in attitudes, polarization, or voting behavior“, which is compatible with previous literature on the impact of bots, and human sharing behaviour. Potential production volume in desinformation says nothing about its impact.
Director Keith Schofield created hundreds of AI-cinema stills for the parallel universe movie ‘Galaxy of Fear’ by David Cronenberg and got himself into a minor outrage cycle on Twitter which is just what the world is today. Some people are upset because AI-art seems unethical, and some are upset because that movie doesn’t exist. How dare he.
Here’s a good thread about how this new artform enables a new way of seeing, and how to approach these strange new worlds with an open mind. And Schofield also did a Sequel: Galaxy of Flemish (1987).
And speaking of AI-cinema: The NYT has an op-ed by Frank Pavich, the director of “Jodorowsky’s Dune”.
I’ve written about AI-cinema extensively here: If Jodorowsky directed TRON, AI Cinema gonna be wild, A selection of fine AI Cinema-stills 1, 2, 3.Ryan Broderick reflects on ten years of This is fine and Yu Jie made a flaming Fine Sweater and published the knitting pattern on Ravelry. Here it is, in all it’s burning metaphorical glory, though i think she should’ve added a ‘This is fine’-speechbubble.
AI-powered "robot" lawyer will be first of its kind to represent defendant in court, in which Do-Not-Pay will give headphones to a lawyer and an AI telling him what to say.
Here’s a paper presenting a technique to enhance voice audio from video input, and basically it can read lips from very low quality and shaky vids. Now you can
totally eavesdrop on anyone you can record on video from miles away, which given the camera abilities shown in Googles Pixel 7 is totally possible, today!
Remember the paper in which they reconstructed sound recordings from videotaping a bag of potato chips and analyzed the vibrations of the plastic surface? It’s like that but on AI-steroids.Social Media Use Is Linked to Brain Changes in Teens, Research Finds: “We can’t make causal claims that social media is changing the brain,” [says] one of the authors of the study. / But I can: Social Media changes the brain. If i'm at my most pessimistic, i'd read that study as a confirmation of my take that socmed established a hormone based attention economy, turning developing brains into dopamine/oxytocin-junkies craving for more. Social Feedback triggers oxytocin shots, and the gamification of socmed provides dopamin. Ofcourse such a media environment has an effect on developing neural circuits.
A new paper on social contagion and the spread of mental health issues via Social Media, mostly among kids. The Guardian has a write up. The topic made rounds last spring with first reports on Tourette-like symptoms spreading via TikTok, a social contagion initiated by influencers with mental health-issues.
Plant Machete by David Bowen: “This installation enables a live plant to control a machete. plant machete has a control system that reads and utilizes the electrical noises found in a live philodendron. The system uses an open source micro-controller connected to the plant to read varying resistance signals across the plant’s leaves. Using custom software, these signals are mapped in real-time to the movements of the joints of the industrial robot holding a machete.”
German audio version of the fantasy classic in 30 parts: J.R.R. Tolkien: Der Herr der Ringe.
Iran Is Using Facial Recognition to Enforce Modesty Laws. Why do you need morality police when you can have face recognition?
Scribepod 1: Some guy has wrote scripts turning Machine Learning-papers into a podcast dialogue for synthetic voices via ChatGPT.
The Creative Underclass is Still Raging: Freddie DeBoer describes how the ‘creative underclass’ is a main factor in online outrage.
AZDNet summarized David “The Hard Problem of Consciousness” Chalmers keynote talk at NeurIPS 2022 in which he predicts that ‘AI could have 20% of sentience in 10 years and the NYT has a piece on the “pursuit of artificial awareness“. Machine Learning Street Talk interviewed Chalmers at that conference, too.
I have a lot of thoughts on this topic, and because consciousness can’t be defined, I’m worried: A machine doesn’t have to gain consciousness or sentience for any deployment of such a machine to be unethical. It just has to gain a machine form of consciousness. We have no clue how that would look like, how that would manifest and how we would recognize it. Just as with hypothetical alien life that goes unrecognized and then contaminated, any AI-consciousness might be too alien for us to detect, simply because we don’t “think” like machines, and machine awareness would be too weird and different from ours. I won’t go full Lemoine on this, but we have no clue what the situation of extremely complex statistical networks look like in 10 years, or 100, and we should be prepared for these questions.Language Models are Drummers: Drum Composition with Natural Language Pre-Training
This British Zoologist Wants to Reinvent Color: “The world’s best colors (...) come not from pigments or dyes, but from materials arranged into crystalline nanostructures that scatter light into 'structural colors'.”
Gizmo has an exzerpt from the book “Robot Ethics” by Mark Coeckelbergh: How Would a Self-Driving Car Handle the Trolley Problem? Ofcourse, there is only one solution to the Trolley Problem.
Russia Reportedly Legalizes Piracy of Games, Movies, and More. Piracy always was tolerated in russia and this is just a propaganda campaign.
Cybercriminals Starting to Use ChatGPT. Ah come on, no way /irony off.
Beware a world where artists are replaced by robots. It's starting now. Molly Crabapple joins the ranks of the AI-art-luddites.
Why writing by hand is still the best way to retain information: I’m not very concerned about ChatGPT for education, because the solution is very simple: Make students write by hand. Sure they can still generate an essay and then copy it, but it would be an obstacle. Also, writing by hand is superior for educational purposes, which is a reason why i’m maybe the only person on this planet who’s in favor of banning all digital screens in the classroom.
Carl Miller - Are You Outraged Yet? Review of Max Fishers book ‘The Chaos Machine: The Inside Story of How Social Media Rewired Our Minds and Our World’.
‘My AI Is Sexually Harassing Me’: Replika Users Say the Chatbot Has Gotten Way Too Horny.
Will We Ever Run Out of Sudoku Puzzles? Nope. “There are 6,670,903,752,021,072,936,960 possible solvable Sudoku grids that yield a unique result (that’s 6 sextillion, 670 quintillion, 903 quadrillion, 752 trillion, 21 billion, 72 million, 936 thousand, 960 in case you were wondering). That's way more than the number of stars in the universe.”
Massive thanks for arranging this list of articles which have each amazed and confounded me