Sustainable Synthetics & Inoculated Minds | PROLOGUE

Towards media ecosystem resilience & mental preparedness ahead of an imminent Generative AI influx

Thomas Cole, Destruction, 1836

Thomas Cole, Course of Empire, 1833-36 (top left - bottom right: The Savage State, The Pastoral State, The Consummation, Destruction, Desolation)

Gestation, Emergence & Aftermath

Before the emergence of Large Language Models (LLMs), diffusion models, and the abundance of creation tools built on top of these, we had to make do with mere glimpses of cultural and media experiments from which to extrapolate how AI might contribute to new forms of storytelling and content experiences in the future.

Experimental narrative games like Agence allowed users to gameplay and commune with responsive AI avatars - this could entail how we engage with our favourite characters from familiar and new stories one day. Japanese idol Hatsune Miko assumed the form of a compassionate AI companion floating in purpose-built hardware by Gatebox - suggesting everyone could experience friendships with their idols attuned to their bespoke needs. The artist Sougwen Chung demonstrated how to co-paint with robots via sensor-laden headbands, hinting that we may eventually be able to direct the stories we create and consume via biofeedback.

Over the last few years, more concrete examples of AI have manifested in the content creation space - albeit limited to experiments due to high costs and a shortage of skills. We’ve seen AI de-age actors (albeit at great expense, like in Martin Scorsese’s The Irishman), co-design synthetic friends to follow (and buy music and merch from, like Lil Miquela), ‘write’ and ‘direct’ film (e.g. Zone Out, 2018).

Then, courtesy of breakthroughs in Natural Language Processing and diffusion, long-established, conceptual visions of automation finally emerged as tangible tools anyone could use for free, courtesy of advanced performance and a UX solve.

If we were to take its lead, AI would enable us to produce any kind of output, irrespective of skill. It would let us write novels, make films (even star in them without ever visiting an actual set), attune stories to our preferences or detected emotional thresholds, and allow us to engage with avatars of newsreaders, therapists, personal trainers, lawyers, doctors, dead philosophers, favourite celebrities or cartoon characters in a highly personal way (possibly even as avatars ourselves). Moreover, my avatar is on the cusp of being able to engage with all of your avatars in social media and vice versa via virtual omnipresence and eventual immortality.

While such possibilities promise a new era for media, culture and the Internet - one that is more DIY, interactive, and personalised - generative AI is no less divisive than the communication and creation technologies that came before it. This time, the pace at which innovations are released into the wild is faster, appeal and adoption are broader, and the financial stakes are higher. The drive to monetise innovations quickly leaves little room for considering their implications or what truly benefits their recipients. Compounding this is a divided industry - while Big AI is coercing the public to want in on their reality-altering productivity revolution, ex-Big AI issued stark warnings about partaking in the beginning of the end. 

This duality raises questions about the impact and implications of the imminent generative AI influx. Can it be reconciled?

Towards Sensemaking

AI’s impact on culture and media has been a theme I have tracked and researched over the last ten years, in addition to studying many other applications of AI and their impact on people’s lives, culture, and society. More recently, a role at a generative AI startup building small language models and tools for the news industry gave me a front-row seat from which to observe a rapidly evolving new market.

As a strategist, my role encompassed making sense of emerging markets, technology trends, and changing consumer needs and behaviours to identify innovation opportunities. It also included anticipating the sociocultural impact of productised technologies to ensure any pursuit of innovation is guided responsibly and sustainably. However, delivering on these responsibilities was challenging due to the unprecedented speed of change - especially against a backdrop of existential dread, limited resources and endless firefighting, both common byproducts of startup life.

It required a sabbatical, mild annoyance at some speculative statistics, and revisiting some paintings from the 19th century to get on top of the sensemaking I felt was essential and urgent. This series is the result of that.

Even before ChatGPT’s launch, the general consensus across the tech and media industries was that generative AI tools emerging throughout 2022 would take the democratisation of content creation to new heights. Industry commentators even went as far as speculating that between 2025-2030, 90-99.9% of all digital content would be synthetic as a consequence (Shoup, 2022; Schick, 2022).

Admittedly, I was bothered by these speculations as they made the rounds. They felt more like tactical media soundbites rather than informed projections, not least because they weren’t underpinned with any actual maths or much else beyond the usual suspects already in circulation (deepfakes, a DIY content revolution, cost savings for Hollywood, mis/disinformation on crack, etc). They did, however, raise some important questions:

What else will synthetic media be - specifically, what media categories will we see evolve and emerge? Who will create it, who for, and with what purposes in mind? Where will it live? Who will care for it and why? Who would get to choose to consume it, and who would have no choice but to?

Then, there was the issue of volume by way of democratisation. The prospect of more content flooding an already saturated media ecosystem reminded me of a painting from Thomas Cole’s Course of Empire series titled Destruction. Depicting an imperial city under siege, crumbling under an invasion of violent conquerors, it seemed like an apt analogy for what could happen to our media ecosystem and its core energy source, attention.

The series, which spanned the “wilderness, growth, magnificence, decline and extinction over centuries of a mighty seat of Empire”, was painted by Cole between 1834-36 as a warning to America at the height of its industrial boom - for “as young and fresh as it was, America was in danger of following the siren call of unbridled expansion” (Riopelle, 2018).

To attract attention, Cole adopted a tactic popular among his contemporaries and actively pursued an emotion-rousing spectacle in his depiction of the siege in Destruction. At the time, deploying this tactic "could be the making of an artist's reputation" (Riopelle, 2018). The irony of Cole deliberately pushing the boundaries to depict an apocalyptic invasion in its savage, stormy splendour to rouse emotion is notable when considering we’ll be exploring tools that could, among other things, take algorithmically-favoured sensation to the next level.

However, in tandem with his intention, Cole’s chosen tactic elevates the series beyond simply being a timeless moral tale. In combining morality and crowd-pleasing sensation, Cole intelligently harnessed the power of culture to make his audience, who either sat at the heart of power and industrial progress or were adjacent to it, aware of the potential consequences of their actions. This makes Course of Empire a blueprint. 

It got me thinking about the role culture could play in shaping the trajectory of generative AI and its impact on the media ecosystem, let alone on our eventual media diets and minds.

Rather than solely relying on Big AI and regulators, how could culture prepare society for what lies ahead and guide the technology towards societal benefit? 

Thomas Cole, Desolation, 1836

The series concludes with a painting called Desolation, which is surprisingly serene and hopeful despite its title and notable lack of humans. Vines climb up a crumbling classical column, now a home to nesting birds, presenting more of a tale of possible regeneration than terminal decline.

This, in turn, revealed an idea the stakeholders in this technology could work towards (specifically, its developers, those deploying it, and the governments and regulatory bodies trying to govern it).

How could the practices of environmental regeneration transfer onto a strained media ecosystem? And rather than getting in the way of regeneration (at worst, causing irreversible damage to the system), how could generative AI be utilised as part of such an effort?

The exploration that ensued to try and answer these questions is captured in this series. I hope you find it interesting - moreover, I hope you reach out with thoughts in the form of ideas, builds or provocations that can make its central thesis better and, ergo, more beneficial to us all (either on here or to lucia@haveanicefuture.com).

NEXT: Introduction