The EU may have reached a provisional agreement on the laws that are going to regulate Artificial Intelligence (AI), but many of us are still left wondering if AI is a powerful transformative tool that will benefit all of us or a social threat that will destroy humanity.
Artist Alan Warburton, who also defines himself as an "AI collaborator", and who is better known for his critical video essays about the impact of technology on visual culture, comes to the rescue with his latest work.
Entitled "The Wizard of AI" and commissioned by the Open Data Institute, the video is a short full-immersion into the complex landscape of generative AI and an exploration of its impact on visual culture and of the fear and loathing associated with this technology.
What makes it remarkable? The fact that, while it contains some archival materials and 1990s clips from a variety of programs, including CNN News and BBC Radio (one of them mentioning the Napster case), the video is mainly composed of stills and moving images generated with the help of Artificial Intelligence.
It took Warburton three weeks to make the video (without AI it would have costed thousands, probably, and he would have needed a proper team of people). More specifically Warburton and his collaborators employed Runway Gen 2 (to generate 16:9 'AI Collaborator' video clips), Midjourney, Stable Diffusion and DALLE 3 (for the still images), Pika (for the fish loops), TikTok (for detective speech synthesis); HeyGen (to generate AI talking detective head), Adobe Photoshop AI (to expand images), Topaz Gigapixel AI (to upscale images), and used Adobe After Effects to collage everything together.
This sickeningly trip through AI or visual treat (depending if you are an enemy or a supporter of AI...), starts with yellow and pink melting monsters, then we discover mutant fish, beautiful fake fashion models and other assorted AI shenanigans.
Warburton defines himself a double agent as he works between tech and art, between theory and practice, and wants to explore both sides of the AI coin, the excitement and the threat or, to put this in his perspective, the wonder and panic – in short, the "Wonderpanic".
This concept that appears at the beginning of the documentary is a feature of modern technocapitalism. Wonderpanic is also the social expression of Bernard Stiegler's theories about innovation intended as a "pharmacological" process.
The French philosopher states indeed that technologies inherently exhibit a pharmacological nature, an idea that implies that each novel technology functions as both a remedy and a poison, prompting the need for additional solutions to mitigate its effects. From this perspective, the introduction of technology initiates a cycle of innovation and adaptation, as society confronts the simultaneous benefits and risks of technological progress. Examples of this process are the almost perfect deepfakes that can be created through these systems, followed by the deepfake detectors that tech companies have also created. All of us - artists, designers, media workers and the general public - are synched with wonderpanic, Warburton tells us.
Warburton then passes to define the concept of "futch", a concept best defined by Joanne McNeil as "every broken promise of every new app or internet service".
Something that is everywhere - in fashion and aesthetics (note this point fashion fans and, in particular, fashion students – wonder if you're even aware of this…), in stocks and shares, and in platforms and devices.
If the technological revolution is stuck in permanent beta, the documentary tells us, AI hype exemplifies this concept: AI embodies the concept of "futch" – it is something revolutionary, tech companies promise us, and it will free us from many burdens. Yet, being unregulated, so far it has just removed a lot of people from their jobs, while at the same time governments all over the world are looking for solutions to regulate it.
One of the most intriguing parts of the video is the arrival on the scene of Midjourney, that, released last November, brought a change of scene. Absorbing millions of images and remixing them, Midjourney created a vast pixel soup, bringing on the scene a new aesthetic drenched in a hyperreal fantasy.
The story is told in the style of a noir graphic novel, Midjourney is the baddie, the hitman who kills the art star: "Turns out the artist had been getting around a little too much, sharing everything – and that's how they got him in the end," says the narrator while an artist lies sprawled on the floor.
The video then makes a revelation that is almost shocking: since mid-2022, AI has produced over 15 billion images, a number that may be shocking, but that it is evident to those of us who are on Discord and see images being churned out every few seconds by the MidjourneyBot on multiple #general channels. "For reference, it took photographers 150 years to produce that many images," Warburton states.
In this immense ocean of synthetic life, Warburton highlights the challenges artists face in navigating the ethical and aesthetic implications of AI-driven creativity.
One on side we know that, being capable of doing in an efficient and fast way the job of a vast team of art directors, generative AI is killing their jobs. Besides, it is also doing so endlessly appropriating, recycling and recombining the work of other artists. That said, generative AI is also producing the best remixed images of mainstream culture, creating a certain aesthetics, a sort of liquid postmodernism, Warburton states.
In defense of AI we may add that certain industries have accustomed us to the remix, to borrow and stealing practices: throughout he decades, techniques employed in music samples were successfully applied to fashion by many contemporary brands to the point that nowadays we aren't even able to say who did that clutch, those shoes or that print first. By now we give for granted and even approve and accept the legal and illegal mutant remixed products manufactured by famous brands; AI does the same thing in a more clever and faster way and without paying for what it samples (in contrast with the practices of the music industry which involve regulation and payment for samples; in this fashion is similar to AI as, quite often designers borrowed from other collections or brands without acknowledging the original). This point would be easily sorted if artists wold be paid for their work used to train an AI, but this seems an impossible task considering the immense appetite of these systems and the super fast pace at which they develop (at the same time this is not a good excuse to avoid paying the artists for their works employed to train an AI).
AI was 2023 word of the year according to the Collins English Dictionary, but AI also influenced the aesthetics of 2023: the video essay highlights this point claiming that "the style of statistical style" became the style. Who knows, maybe that's why we are left with unstable fashion collections that seem to embody the concept of "futch" with designers incapable of producing timeless designs (will we ever have the next Alaïa if we are trapped into the futch?).
As Warburton touches on issues such as intellectual property, the changing dynamics of visual culture, the essence of art in this evolving landscape and the potential directions AI art might take in the future, he is faced with more questions and dilemmas, "Has the apprentice grown too strong? Are we Oppenheimer, Turing or Obi Wan? Are we training genies for Thanos? Is photography a zombie?" Warburton wonders towards the end, closing with a more optimistic note, "it'll probably be fine, artists will find a way. And in the meantime, show me some AI candy!"
The video essay is not without controversy, though: to avoid stealing Warburton fed his own art to generate with Midjourney the sections he needed, yet in this way he trained the system and helped it improving and learning. Still this is his revenge: most articles about AI focus on companies and big players, OpenAI's Sam Altman is often in the news, for example, but those workers whose jobs were impacted by AI aren't.
Combining critical analysis, educational insights, and a touch of satire, the video essay closes with a bitter note: in the titles there is an "In Memoriam" section that doesn't include names of collaborators sadly departed while doing the video, but of a selection of figures that may be killed (or have already been killed…) by AI, including voice artists, photographers, compositors, screenwriters, set designers, architects and more.
Yet while asserting the death of these figures Warburton actually shows that AI still needs human interaction to achieve the most perfect results. "The Wizard of AI", the world's first ever AI documentary, wouldn't indeed have been made without clever and skilled humans producing prompts and pushing systems to reiterate certain images (yes, it does take a long time to obtain what you want and, if you're not careful, AI will easily trick you making you believe you have created something unique while that's not the case…) and without a strong text written by Warburton himself, based on research he has conducted for his ongoing PhD in Digital Culture and Communication at Birkbeck's Vasari Centre in London. In a nutshell, yes, AI may be fast and capable of generating the most visually striking and extraordinary worlds, but, for the time being, it still needs talented humans to function at its best.
The Wizard of AI from Open Data Institute on Vimeo.
