6 minute read
NEXTGENERATION ART
The Risks of Generating Creative Content with Artificial Intelligence
There is a distinction between creative pursuits and formulaic ones.
A task with variable outcomes and subjective elements, or one which requires novelty, innovation, and expression is usually considered creative. If it’s repetitive, standardized, or predictable, then we call it formulaic.
It’s the difference between sculpting a work of art out of clay and assembling a shelf from IKEA. It’s actually less of a binary distinction: most of the work that we do lies somewhere on the spectrum between creative and formulaic.
Now, when we imagine how Artificial Intelligence (AI) will impact us, our thoughts first go to our work. Automation is often discussed; we’re naturally invested in how AI will alleviate the burden of completing formulaic tasks and free us up to focus on more creative endeavors.
But as we actually enter ‘The Intelligence Age’ it seems like we’ve got it backwards: AI is instead excelling at creative tasks. Take the AI system DALL·E 2, which creates professional-grade images based on a worded description, or ChatGPT, which can write short stories in seconds. In light of this, we should be considering the unique risks of generating creative content with AI, and what it means to automate creative tasks.
To understand this better, I think it's worth looking at the field of generative art. It’s a captivating form of artistic expression because unlike traditional art, which relies on the artist's direct input and intentional decision-making at every stage, generative art embraces the power of algorithms and rules to drive the creative process.
The artist does not actually hold the brush so-tospeak; they define how the brush can move, and they let the system do the painting (also, it’s worth noting that within the genre of generative art lies the subgenre of AI-generated art).
The process of creating generative art is naturally more computational and digitized than with traditional art, and so too is the means by which it’s consumed, experienced, and purchased.
A brief aside: generative art was at the center of the ‘web3’ decentralized ownership movement that hit fever pitch in 2021-2022, which is itself the subject of controversy.
Generative art naturally lent itself to being monetised and traded on the blockchain as it was already created algorithmically, and an enormous amount of computational power was used to record who owned what piece of digital art. This translated to energy used and carbon emitted; each time a piece of art on the blockchain changed hands on the Ethereum blockchain it produced 103.42 kilograms of carbon dioxide (in 2021).
Notably, subsequent upgrades to the underlying technology have reduced this carbon output drastically, and a single transaction on the Ethereum blockchain now produces an impressive 18.72 times less carbon than a single Mastercard transaction.
But the carbon cost was not the only cost — the scene rapidly became a home for grifters, peddlers, and fraudsters due to the anonymity and lack of regulation around the blockchain.
Molly White’s website web3isgoinggreat.com estimates the running total amount of money lost so far to blockchain grifts and scams at 12.182 billion US dollars.
The important lesson here is that when we use computational techniques to digitize products and processes we fundamentally change the way in which people and other digital systems can interact with those products and processes. Generating art with computers allowed for easy integration with blockchain technology, and for people to commodify and trade art in new ways, and at unprecedented scales.
The effect of this is still being understood, but as mentioned, there were clear negative outcomes.
We’re now shifting towards generating not just art, but general creative content (e.g. images, blog posts, tweets, emails, reports, essays, etc.) using AI techniques. And technologies that allow for the monetisation of this kind of content have already existed for the past few years.
One broad and pessimistic prediction for the future is that the ability to rapidly manufacture creative content using AI — in tandem with the incentive to generate attentiongrabbing content for the purpose of monetisation via advertising-based business models — might contribute further to the commodification and fractionalisation of our attention spans.
But that’s only a prediction. It’s unclear at best what will truly happen in the coming years.
However, if we consider this new paradigm of AI-generated creative content analogous to how generative and blockchain technologies affected the art scene, then I feel we can at least move forward knowing that the following will be important:
Implement regulation and safeguards - Establish rules and measures to protect users and prevent scams. If the best forms of creative content (e.g. educational content, essays, journalism) can be scaled up, so too can the worst forms of content (e.g. email scams).
Promote education and awareness - This is key with any new piece of technology. Direct efforts must be made to educate people about the implications, benefits, and risks of AI-generated creative content.
Closely monitor how technology affects people - How we use AI needs to align with ethical standards and societal well-being.
In particular, we should be identifying economic and labor implications of AI, assessing the psychological wellbeing of users & those impacted, and understanding how AI generated content is impacting social dynamics.
This leads into the last point, which is to…
Continuously evaluate and adapt - Stay updated, adjust strategies, and update practices both individually and collectively as we learn more about the pros and cons of AI-generated creative content.
ISAAC R. WARD
applies Artificial Intelligence to the diagnosis of cardiovascular diseases, whilst also solving technical challenges in the field of Computer Science. Specifically, he is interested in how these algorithms learn to mimic complex human capabilities.
As an advocate of using AI within the medical industry, Isaac is also interested in leveraging these technologies to improve the quality and accessibility of healthcare.
He used his 2020 Fulbright Future Scholarship to pursue these interests via postgraduate studies at the University of Southern California, whilst also exploring how AI research is rapidly applied to industry challenges within the United States.