The Age of Hyper-fans, Part 7: The Return of Dystopias: Why Generative AI Won’t Destroy a Single Job
People are worried about losing control of their finances; people are worried about the meaning of their work; people are worried about Artificial intelligence (“AI”). This worry, and it should not surprise you if you read this far, is one and the same (See Part 2).
In recent literature — and more broadly in the Twentieth century — governments are not held in a high regard. It is telling that the term Dystopia was coined by John Stuart Mill, a liberal writer, in 1868. Dystopia, Utopia and science have always been closely related. In Men Like Gods (1923), HG Wells describes a world that he calls, in a somewhat uninspired fashion, Utopia. In this world, technology has enabled a post-government collaboration between humans (Web3, anyone?). Meanwhile, Brave New World (1932) depicts a sanitised dystopian world where technology has impeded the uniqueness of human emotions and spontaneity.
Influenced by the Second World War, later dystopias were less focussed on technology itself and more on the dangers of fascism and totalitarian forms of government. Totalitarian governments, however, are, by design, enabled by advanced technology. That is because totalitarianism is not just a form of dictatorship. It is a system where the private and public speres, to borrow Jürgen Habermas’ concepts, are completely merged until the individual cannot hold any secret from the State — “Secrets are Lies”, says the fictious equivalent of Google in The Circle. (Why large corporations are akin to states in recent literature is explored in Part 4).
Totalitarian regimes need mass surveillance — which is, of course, enabled by technology at a scale far larger than was possible for most of history. This is evident in many works of literature where mass surveillance is a common State activity, but employed in very distinct ways today compared to the Cold War era. For example, spies as a means of worrying State surveillance have disappeared, transformed instead into plot tropes for light-hearted action and comedy (see the numerous, humorous and dubious “young spies” films such as Cody Banks and Alex Rider).
The debate around whether technology will lead us to utopia or dystopia has resurfaced in the light of recent developments around AI. Le Monde, an established French newspaper, recently published a piece called “Behind AI, the Return of Technological Utopias”, quoting, amongst others, Sam Altman’s concept of “techno-optimism”.
Regulators, journalists and theorists’ concerns are evidently literature-inspired. The EU’s AI Act mentions it is “human-centric” (as opposed to rabbit-centric I suppose). Reading the uses that are prohibited by AI, one can find heavy references to mass surveillance. For example, facial recognition or biometrics data are mentioned several times.
The internet is full of articles raising alarm bells. My personal favourite is a piece called “How AI Will Go Out Of Control According To 52 Experts”, a truly mesmerising combination of clickbait and uninspired quotes.
Behind these fears, again and again, we find references to a lack of control and transparency, which, as we have seen, is something AI shares with finances and food. Why we feel this way is explored in Part 2 and 3.
The danger with this thinking is that it frames AI as something that should always perform perfectly. That is true, when there is no human validation and stakes are high (like in healthcare). We should however remember that humans, themselves, do not perform all that well, most of the time. We get things wrong, mess up facts, are inconsistent and distracted. In fact, AI could be a good tool to reduce biases.
Because younger readers are a bit less worried about surveillance, recent works of literature are rather ambivalent on technology, reflecting perhaps a more balanced perception of its benefits and dangers. In Ready Player One, the hero realises that he must get out of his metaverse-like universe to truly forge bonds; but its sequel, Ready Player Two, depicts a rather awkward ending where characters upload their conscientiousness to a data centre to colonise new worlds, which is, somehow… a good thing? All this because the AI copy of a dead character will not die, and that is, for the characters at least, fine.
This particular turn of events shines a light on another characteristic shared by the most extreme tech utopians: the promise of immortality, of which much has been written.
In a different vein, some works have argued that young adult literature depicts worlds in which technology can actually improve the agency of heroes; by providing tools that enhance people’s sense of self-worth and identity (for example, by removing the constraints of bodies on gender identities).
This debate is more philosophical than practical, and in practice it turns out the transhumanist movement, which aims at enhancing humans by means such as eradicating diseases, haven’t had much progress in the past ten years. As Zoltan Istvan — a person whose job is essentially to be an influencer for transhumanism — says himself: “The longevity movement has been disappointing in what it’s produced; there was a lot of hype….On the other hand, I have to say, AI, […] has gone faster than most people realize. […] I’d say 10 years ago, I wouldn’t have guessed that we would be this far along where I might be able to write half my essays using a chatbot for free.”
Istvan says something of importance for the purpose of this analysis. The irruption of “for free” is a telling argument to the true nature of the AI evolution (for ChatGPT isn’t a revolution, but the fine-tuned version of GPT-3, a large language model that already performed very well back in 2020).
Business revolutions are always cost revolutions — from the steam powered machinery that automated the textile industry to the internet that reduced cost of distribution to zero. Just like our economies are still running at historically low level of unemployment, employment opportunities tripled in the textile industry in Britain after the introduction of spinning-machines.
Generative AI is another business revolution; and in line with past ones, it has to do with costs: it reduced the cost of content creation to zero. Which means, combined with the internet, that we live in a world where content creation cost is zero and content distribution cost is zero.
The most advanced take of what the internet has changed for costs has been theorised by Ben Thomson on Stratechery, with Aggregation Theory, which frames the aggregation (a prerequisite to commoditisation) of content and service providers by internet companies.
In summary, and in Thomson’s words: “Zero distribution costs. Zero marginal costs. Zero transaction costs. This is what the Internet enables, and it is completely transforming not just technology companies but companies in every single industry. Old moats are gone — and new ones can be built.”
As with the internet, Generative AI will have profound, cost-driven changes on businesses.
Does this mean the era of “Content is King” is over? To the contrary. People want more content — the demand is insatiable. For universes and stories that work, though, it means an era of unprecedented content generation. No longer is JK Rowling needed to write JK Rowling-like stories; a storyboard suffices, and her style can be imitated (to an extent, of course — we explore how this could work in Part 8).
There remains, however, space for human-written content, for two reasons.
First, because AI is statistics and cannot reproduce the sheer brilliance of stories and novels. These winning stories, though, will be more and more monopolistic in nature as they take an increasing share of our attention.
Second, because there is a common misconception that content production was at a market equilibrium of sorts, whereby demand for content met the offer of content. That is not true: content creation was always a bottleneck. Not for all content; but for hyper-fan compatible content. In other words: there is only to imagine a good story, the rest will follow.
I wager that this evolution will be an accelerator for the expansion of content universes. Rather than chunk through swathes of random new content, we will live in content spheres of influence, within each sphere a community.
What about businesses themselves?
In this matter, I expect to see the “Powerpoint effect”. In practice: there were very few slides produced before Microsoft acquired Powerpoint (then called Presenter). Now, about 30 million Powerpoint presentations are created each day. Was there a pressing unmet need for Powerpoint presentations? Probably not. There was, however, a pressing need for communication and documentation, of which we will produce more; simply because now we can.
What about blue-collar services, such as call centres? Contact, too, is a bottleneck. The market for Contact Centres is expected to keep growing at a fast pace, from $340bn in 2020 to $500bn in 2027. And contact centres have not been waiting for Generative AI to innovate — they have been at it for years. Their reason is simple: hiring is hard and a lot more is expected of them as digital services grow. Post pandemic, these problems have compounded, and this translated into the tripling of wait times in call centres. This is even though AI offerings based on GPT-3 or well-performing NLP algorithms have been available for years.
For the content creators and journalists, AI will be a tool more than a threat. Low-level journalists have already been replaced by clickbait generating bots, and reporters and investigative journalists do so much more than write at their desks.
In short, works calls work, and when things get cheaper, we just produce more — or, rather, abyssus abyssum invocat.
Want to read more? Stay tuned for the final part of the Age of Hyper-fans, Part 8: Hello, World©: The Curious Case of Software as Content
[1] “ Contrairement à une idée largement répandue, la langouste se nourrit exclusivement de fruits de mer. Ce qui ne l’empêche pas de rester très humaine. “
