Writers, designers, and artists are not exactly enamoured by the rapid developments in AI ‘creativity’. Within the space of a couple of years, AI-powered generators have evolved from complex research tools into free-for-all search engines of the unknown. Today, anyone can enter a set of parameters and watch as computer-generated text and images are returned, free from copyright, plagiarism, and precedent, and perhaps even containing more than a germ of original thought.
AI is effectively a mirror, sifting through the vast database of human creativity and cannily blending a bit of this and that in order to return a visual representation approaching or transcending that which we’d imagined, or maybe a chunk of text that hopefully organises a coherent set of thoughts. It comes with inbuilt biases gleaned from the source material and still suffers the occasional baffling but telling lapse of judgement or coherence.
ChatGPT, the AI writing bot
Nevertheless, it’s getting better. It took about ten seconds for ChatGPT to turn a 14-word prompt (‘Write an article about artificial intelligence and imagery in the style of Wallpaper* magazine’) into a 300-word ‘article’, coherent and credible enough to pass muster for anyone skimming the site for a primer on the topic or a bit of background. Take the following paragraph:
‘One of the most exciting applications of AI and imagery is in the field of visual recognition. With the ability to analyse and interpret vast amounts of visual data, AI algorithms can be trained to recognize and classify objects and scenes with incredible accuracy. This has a wide range of potential applications, from helping robots navigate complex environments to identifying and tracking objects in surveillance footage.’
There’s nothing inherently contentious there, let alone any real indication that it was written by either a machine or a human. Admittedly, the text is lacking in zip and is only marginally duller to read than if one had to actually write it. But in the fast-paced and algorithmically driven world of online journalism, it might just suffice, all the more so if an AI could be tapped up for the latest set of keywords and trends. Better still, those keywords could be automatically updated within the body of the text, ensuring an article was always bobbing about near the surface of the flotsam covered data ocean.
For writers, journalists, editors, and other content creators, the future looks bleak. How Google and others respond to the challenge of automated content remains to be seen, if it even regards it as a challenge worth tackling (unsurprisingly, Google invests substantially in AI, including ChatGPT’s creator, OpenAI). Most people remain blissfully ignorant of the hidden mechanics of search, so a fresh flood of AI-generated data is unlikely to darken the murky pool of references and recommendations any further, at least for now.
AI’s unseen roles
From our conversations with tech leaders, it seems the deployment of AI starts with things like image processing, shoring up backdrops and compression artefacts on video calls, tweaking audio to mute background noises, etc. These largely invisible processes will likely remain cloaked, even as AI continues to posit more ethical questions than it can conceivably answer.
AI imagery: transforming architecture, autos, art
From a creative point of view, there’s a lot more that can be done with the visual tools. Architects and artists are already deep into the potential of AI portals like DALL·E, Stable Diffusion and Midjourney, writing ever-more sophisticated prompts that unearth strange new aesthetics and juxtapositions. American architect Andrew Kudless’ Matsys Design (opens in new tab) and the Iranian architect Mohamad Rasoul Moosapour (opens in new tab) are just two of the studios using AI prompts to push for new forms and genres, while commercial tools like ARCHITEChTURES (opens in new tab) promise to turn these capabilities into ways of shaping real-world structures. They invite us to imagine a future that blends computer-aided production and construction with AI-guided parametric design, moving the bar just a little bit further than even the glossiest CGI render.
Visual disciplines benefit when the brief shifts from imitation to creation. Automotive AI (opens in new tab) showcases cars that never were, heaping on the clichés yet still forging an original path, while An Improbable Future (opens in new tab) makes masterfully credible mash-ups of conceptual machines, conjured up from online traces of Dieter Rams, Sony and Bertone.
Photographer Mathieu Stern (opens in new tab) splices vintage cameras with art history to create an alternative timeline of photographic technology. The Guardian recently invited six contemporary artists (opens in new tab) to dabble with AI-generated imagery, and published the resulting blend of clipped images, approximations, and guesstimates. Depending on how specific and related the prompt was to the artist’s current practice, the results verged from random guesswork to potential usable sketches, research, or even finished artworks. The field of generative art, as surveyed in Wallpaper’s January 2023 issue, treats AI as just another brush and palette, along with open-source coding tools, NFT distribution, and more ‘traditional’ digital methods. As with all artistic practice, intent is key.
Baked-in bias and limitations
On the flipside, there are the implicit biases baked into the system. Lensa’s sub-Loaded preconception of what a female-representing self-portrait should be. A number of artists have actively rejected the technology, citing its ability to mine existing databases of imagery and effectively ‘take over’ a particular aesthetic (artist Greg Rutkowski’s elaborate fantasy landscapes (opens in new tab), for example). It’s also a window into our limitations. Sites like This House Does Not Exist (opens in new tab) serve up an infinite diet of click-bait-worthy shelter porn, trained on decades of idealised, utopian imagery generated by the industry itself and published by the likes of yours truly.
The human touch
Architects and magazines are being trolled by the predictability of their past offerings. To go further requires specific human intervention and direction.
Ultimately, the question is how concerned we should be about the plethora of computer-generated creativity. Purely digital creations are no longer noteworthy today, whether it’s a ‘virtual’ K-pop girl group like Eternity, a blockbuster fashioned entirely from CGI, a piece of loot or a skin from a video game, or even, heaven forbid, an NFT. Will these things continue to be celebrated when the (vital) human component is removed from their creation?
For now, what keeps this swirling vortex of confusion honest is the art of curation, or as it used to be known, editing. Here at Wallpaper*, the mantra ‘it’s all in the edit’ has been a guiding principle since the earliest days of the magazine. The internet ensured the firehose of cultural production was kept at maximum volume, but it’s our task to sift through the output and find the things that transcend Twitter spikes and trend analysis. With AI, that task is greatly expanded, but also more focused, as the common threads and trends that can be so efficiently mapped and drawn out by machines swiftly rise to the surface.
ChatGPT, write an article about artificial intelligence and imagery in the style of Wallpaper* magazine: ‘Overall, the use of artificial intelligence in the creation of imagery for Wallpaper* magazine has significantly improved the speed and efficiency of the design process. It has also allowed designers to push the boundaries of what is possible in terms of visual creativity, resulting in truly stunning images that captivate and inspire.’
We can still do much better than that.
ChatGPT (opens in new tab)
DALL·E 2 (opens in new tab)
Midjourney (opens in new tab)
#Artificial #Intelligence #writing #wall #creative #professions