Want to know what it takes to make a hit movie? Ask a bank analyst.
That’s one of the takeaways from Signal In the Noise, a compendium of deep thoughts released annually by the media and telecoms team at Barclays. The 2023 edition weighs in at 133 pages and includes lots about how artificial intelligence might affect the creative process. Edgar Degas, Vincent van Gogh and Aaron Hertzmann are all quoted within the first few pages of the first essay, which ruminates at length on the nature of art:
In some ways, the unpredictability of the experience, the complete surrender to the artist is the point of art — whether it is the worlds imagined by James Cameron in Avatar or simple strokes of a Picasso. That surrender is not possible with technology alone, because technology is context independent and is merely a tool. Art, on the other hand, “ . . . requires human intent”. Technology in itself independent of the artist does not really have intent.
Art and content is about the interaction between the context of the creator and the consumer, which is why the same work can move people differently. This is why art is never “complete” and decoupled from a creator, it cannot really morph contextually.
And so on.
One theme that emerges is the difference (in monetisation terms) between big-budget video and more passively consumed media like streamed music. It’s a distinction at the crux of Chapter 3 — “Why Hollywood writers misunderstand the role of AI in content creation” — and it’s timely stuff. Hollywood studios yesterday reached a tentative deal with the Directors Guild of America union to avert a writers’ strike shutdown. Their agreement includes a clause that “generative AI cannot replace the duties performed by members”.
Pah! says Barclays; a point it illustrates by comparing writers’ rooms with the second industrial revolution:
While understandable, this opposition, in our opinion, represents a fundamental misinterpretation of the role of technology in content creation. In fact, the impact of AI may be the opposite of what content creators seem to be implicitly assuming, in terms of relative power balance between studios and creators.
Technology tends to fundamentally restructure pre-existing workflows into different forms, which ultimately results in new capabilities. These changes are often not synchronous with the emergence of new technology, one of the factors that often gets missed during the hype cycle associated with new technology. For instance, when the world shifted away from steam engines to electricity, productivity levels counterintuitively dropped for decades. Based on the evolution of the industrial economy over the last 100+ years, when society sees big shifts in general purpose technologies like that away from steam towards electricity and now the shift toward AI, the propensity tends to be to plug the innovation into an existing business process. Steam-powered factories, for instance, were designed broadly based on a central engine room which then was connected an axle to other parts of the factory to drive various equipment. However, due to constraints around how large these axles could be, most of the factory was clustered in layers around the central steam engine. During the early days of electricity, factories replaced the steam engine with an electric motor but kept the rest of the factory layout the same, which limited the productivity benefits of using electricity. Over 30 years, when the ‘electricity first’ generation gradually came to replace the older generation, factory layouts were fundamentally redesigned to use the decentralization allowed by electricity which then drove up productivity and altered the manufacturing paradigm.
And so on.
The purpose of the above analogy is to suggest that a company’s instinct, when presented with any new technology, is to cut costs. Possibilities may become radical changes in the long term but redesigning workflows is costly, so would undo the quick-hit benefit.
Also, industrial revolutions don’t really apply to media because “technology and capital don’t necessarily enable more scale,” Barclays continues. “For instance, does it really make sense for Disney to make more content simply because it can do so more efficiently with AI?”
🚨🚨 McLuhan klaxon:
In order for technology to impact scale, it has to alter the nature of content itself. changing the nature of content rather than merely enabling higher volume of content. Photography was not about replacing paintings as much as creating a new art form. [ . . . ]
The assumption today, as evident in the demands put forth by the Writers Guild, is that capital controls technology’s intent independent of the creator, which we find to be a highly reductive argument. We note that Ted Chiang, one of the most highly regarded contemporary science fiction writers, frames AI as way to extend capital’s control over labor.
In our view, this is not uniformly true of every technology and of every use case. Even if AI is able to write passable stories, as Hollywood writers seem to fear, it is not clear what problem that is really solving for. Netflix likely receives thousands of scripts every year, and it is not clear why the company will need to lean on AI to create stories when it probably cannot process even a fraction of what it receives today
And so on.
What follows is another section on the nature of art, this one specific to how it reflects society and culture.
Human artists tend to repurpose existing material based on whatever is current — the examples Barclays chooses include Kurosawa using Shakespeare plots and the cross-pollination of ideas between The Beatles and The Beach Boys — but what if AI can throw in a curveball? AlphaGo beat Lee Sedol at Go on their first game by playing a completely alien move, so maybe it can do the same for sitcoms and true-crime podcasts?
Viewed from this perspective, the role of AI could in theory be to enhance and accelerate creativity by potentially allowing creators to see numerous iterations on demand without a long drawn-out tail and error process. While this could be more about increasing productivity in the short term, over the long term, this could drive the creation of new art forms that wouldn’t be possible without AI.
Given the above, if anything, AI should in theory be used by the best content creators to become even better and tougher to replace, as technology on its own doesn’t have intent, independent of the creator. The best talent will become even more valuable as technology will enable further enhancement of their work the same way Walt Disney or James Cameron incorporate technology to widen the canvas for storytelling.
Did that sound like a conclusion? It’s not. The team has more to share on the nature of art, this time reflecting on the theme of concept versus execution.
In essence, AI will make it cheaper and easier for amateurs to make stuff that looks and sounds professional. That’ll mean more stuff. But that’s OK, because the professionals will remain better at making money from stuff.
Talent that is good will have more ways to express itself without a lot of capital but capital doesn’t automatically get any scale advantage from technology. In fact, the quality gap between Disney and the average amateur creator may be narrower because of technology.
However, it is interesting to note that in this age of content fragmentation, Disney and Netflix have actually cut down the amount of content they support, and have grown their shares in their relevant content categories. Similarly, the lifetime value of music stars like Taylor Swift, Beyoncé and Drake is higher than any other stars anytime in history despite the number of tracks of streaming crossing 100mm. Scale for a company like Disney comes from taking every piece of content and maximizing its yield while scale for YouTube comes for the fat tail. The latter depends on broad tech tool access; the former has very little to do with tech and in fact, tech has very little if any impact on maximizing yields for Disney. AI may not change Disney’s content volume but Disney may be able to make its content better.
The reason both these ecosystems (user-generated and premium content) remain non-overlapping is because volume explosion typically gives rise to new distribution channels like social media, YouTube and TikTok that don’t necessarily overlap with the channels for other content forms. Creation technology evolution is normally paired with distribution technology evolution as well and as of now, there is barely any discussion of the latter.
And so on.
“We believe the present debate is highly reductive and, if anything, the impact of this technology is likely asymmetrically skewed toward creators rather than media companies, contrary to the popular narrative,” Barclays concludes, for real this time.
Clients get no specific advice on how to trade this brave new media landscape — though a few pages later there’s a long chapter called “Why NFTs will change entertainment”, so at least they get that.
— Fourteen reasons why AI pop won’t eat itself (FTAV)
— Disney and the death of cultural transmission (FTAV)