Embark on Sora’s Odyssey – The Newest Era of AI Artistry has Begun

By Martijn de Boer

“A stylish woman walks down a Tokyo street filled with warm glowing neon and animated city signage. She wears a black leather jacket, a long red dress, and black boots, and carries a black purse. She wears sunglasses and red lipstick. She walks confidently and casually. The street is damp and reflective, creating a mirror effect of the colorful lights. Many pedestrians walk about,” reads the prompt of one of Sora’s showcased, seemingly lifelike videos. 

Artificial intelligence, and especially generative AI, has taken the world by storm in recent years. ChatGPT, for one, has become a household name since its introduction in November 2022. As a result of its unprecedented accessibility and efficiency, anyone with access to a laptop or mobile device can now use AI to aid them in their creative pursuits. Ranging from writing entire essays in the blink of an eye to crafting artworks in the style of Monet or Dali, an algorithm can create whatever the user desires based on a simple prompt. A new digital age is heralded by generative AI, and the possibilities are endless. 

Enter Sora, the latest installment in the OpenAI-lead generative AI frontier: a cutting-edge AI tool curated to video editing. Using the same deep-learning techniques as its predecessors, it allows for users to create lifelike videos by way of a simple prompt. Be it an illustrious overhead shot of a snowy afternoon in Tokyo, or a dalmatian traversing across balconies in a colorful Italian village, Sora opens up the possibility for your everyday user to become a filmmaker in their own regard (which is especially topical now in the advent of the Oscars) – in turn augmenting their creative vision and expanding the horizons of their artistic potential.

While the promise of generative AI offers limitless opportunity to explore thus far uncharted realms of artistry, previously confined by the bounds of imagination and talent, one has to be wary of the potential impact this new and highly disruptive technology poses on the creative industry altogether. Generative AI services showcase incredible capabilities, and their ensuing creations often blur the lines between human and machine-generated content. Yet, artificial intelligence is in itself not a creative entity, and its training necessitates the sourcing of millions of datapoints by way of scraping the internet. Only through this comprehensive process, the AI tool contains the magnitudes of data needed to create artworks, articles, songs, or videos that bear a seemingly human-touch, by virtue of its algorithmic prowess. Exactly therein lies the relevancy from the creative and copyright-oriented perspective: how can a generative AI model create a painting in the style of Monet if it does not already have access to the artist’s works? 

The same critical tenet can be applied to the generated work in its own regard: should a user be granted authorship of the created work if they simply used an AI tool to create it? The consensus seems to point to the negative, considering the insufficient amount of human creativity involved in the overall process. Yet, legal opinions diverge on this topic, as the Beijing Internet Court recently ruled it is warranted to bestow copyright ownership onto the person who prompted an AI-generated image. Notwithstanding the possibility, the question remains whether it should, and whether such an allocation is a desirable outcome. The entire notion of generative AI stands at odds with the ratio of intellectual property law: traditionally, innovation and creation beget protection by virtue of the substantial degree of effort that goes into the creative process. The use of an efficient, low-stakes tool like generative AI ought not to gain the same economic benefits then, one could argue.  

Glaringly, the aforementioned scenarios are largely incongruent with the realm of intellectual property law, and provoke questions regarding ownership rights and alternative rights-clearance mechanisms (for example, licensing). Yet, due to the high degree of legal ambiguity and internal obscurity – OpenAI has shown to be highly reluctant in divulging information regarding the datasets its models are based on – there is little substantial basis to initiate action on. The lack of a clear legal framework complicates things further. Whereas the EU’s AI Act and Biden’s Executive Order on AI show great promise and are highly anticipated, these legislative acts will nevertheless take significant time before they are effectively and practically implemented.

In the meantime, ownership disputes run rampant: in December 2023, the New York Times sued OpenAI for copyright infringement on millions of the esteemed newspaper’s articles. OpenAI, in return, falls back on the fair use exception under the US copyright regime. Pending judgment, and awaiting future legislation, it is up to the collaborative effort between policymakers, industry leaders, researchers, and developers alike, to foster a responsible and inclusive landscape – one that encourages creativity and progress, yet addresses adequately the potential legal, ethical, and social implications of generative AI. This tension – and this changing digital landscape (think of, for instance, the advent of Google) – is not a new phenomenon, neither is ultimately the necessity for copyright regimes to adapt to technological innovation. The question then lies more so in how this adaptation should take place in light of the remarkably disruptive nature of generative AI.

If one were to zoom out from a pure IP-perspective and bring to light the potential ethical concerns surrounding generative AI, the situation becomes all the more dire. One of the most pressing concerns currently is the proliferation of misinformation and manipulation facilitated by generative AI tools – something that is especially relevant when it comes to Sora. A state-of-the-art deep learning technique that analyzes and manipulates video content sounds incredible. Yet, the possibility of creating highly realistic but fabricated and harmful videos remains a serious danger as a result of the technology’s widespread access. The generation and dissemination of false information have evolved significantly, making it even easier for public opinion to be manipulated and polarized, and facilitating the unprecedented defamation and misrepresentation of any individual. 

In late January 2024, explicit AI-generated deepfake images of Taylor Swift made the rounds across social media to the point The White House officially commented on it and Congress initiated action. Due to the lack of transparency surrounding the technology’s inner workings, tracing the origin of the controversial posts proved especially cumbersome, while they accumulated millions of views in the meantime. Despite rebuttals from both OpenAI and Microsoft regarding their responsibility and respective content prohibition policies, the highly publicized situation underscored all the more the importance of concrete internal safeguards, ethical guidelines, and regulatory frameworks in curtailing the unrestricted creation of misleading and malicious content, and holding the appropriate parties accountable.

Whereas potential misuse was previously stymied by a keen eye and the model’s inability to imitate reality perfectly – artworks created by generative AI generally appear somewhat dreamy, or feature anatomically incorrect bodies – artificial intelligence has since witnessed exponential growth in both accuracy and sophistication. In just a few years’ time, profound technological advancements have enabled generative AI services to create increasingly sophisticated and convincing content across different modalities. Sora is no exception: whereas a video of two dogs hosting a podcast in the mountains hinges (only ever so slightly) on the surreal, the announced capabilities of the text-to-video model are a testament to the technological prowess it possesses. 

Navigating the myriad of potential challenges in the shrouded voyage of generative AI requires a multifaceted approach: one which involves a sufficiently tailored protection that is aligned with the interests of all parties involved. Legislators should strive to achieve a fair balance between safeguarding the creative spirit and fostering highly promising technological development. In a similar fashion, while generative AI serves as an immense catalyst for creativity and progress, developers of AI-driven models like Sora must implement robust mechanisms to detect and prevent misuse. Ultimately, promoting digital literacy and a sharp degree of critical thinking can enable individuals to discern between authentic, AI-generated, and AI-manipulated content, mitigating the potential detrimental impact of misinformation on society – something that becomes more difficult to accomplish as each day passes, and each new cutting-edge AI tool is introduced to the public. It remains to be seen whether adequate measures will be taken to combat the variety of pressing legal and ethical concerns; or, whether the world will turn a blind eye to them, favoring in lieu of such a future in which human ingenuity and artificial intelligence converge to redefine the boundaries of creative expression.