AI Insights: Should we be worried about Sora?

Should we be worried about Sora and other AI tools?

In a world captivated by the mesmerizing capabilities of Sora, the notion that the future may not require artists is stirring conversations. With the ability to craft up to 60 seconds of highly detailed and exhilarating video with complex camera movements, Sora introduces a new era of AI for video. The concerns voiced by artists echo fears of careers being overshadowed by AI and the looming concern of deep fakes. Let’s dissect…

The future will still need artists

Let’s acknowledge the perennial struggle of artists. Even my husband, an accomplished photographer who has captured moments with several presidents, faces the challenge of a shifting industry. Professional photos often lose ground to snapshots taken on iPhones. The age-old narrative of the struggling artist persists.

Despite Sora’s remarkable prowess, the challenge remains for artists to bring forth the precise images lingering in their minds. While Sora’s examples are undeniably impressive, the question persists: do they align with the artist’s vision, or are they the result of settling for what the AI deems “Good enough?” In my venture last Halloween, attempting to conjure ghosts, 99% of the AI-generated results proved unusable, even with refined prompts. AI, it seems, remains unpredictable.

I foresee a future where AI becomes a tool for artists rather than a replacement. Artists will evolve into prompting experts, mastering the art of generating aesthetics through precise prompts. Consider a film director employing AI in their project – they may enlist a prompting expert well-versed in filmmaking, art, lighting, and more. This expert, in turn, collaborates with traditional artists to rectify the nuances that AI will miss.

Copyright: Who Owns AI Artwork?

The US Copyright Office has ruled that AI-generated art without human input cannot be copyrighted under US law^[1]. However, the US Copyright Office released newer guidelines, stating that copyright will protect the human-authored aspects of a work that contains AI-generated material, which is independent of and does not affect the copyright status of the AI-generated material itself.^[2] This new ruling came after an artist used Midjourney for a comic book. A copyright was granted for individual images in the book, but not the entire book.

On the flip side, AI developers trained generative AI on copyright-protected art without permission or compensation to the artist. We’ve seen several class action lawsuits for copyright infringement from artists who claim that AI tools like Midjourney and Stabile Diffusion violated their copyright because the AI was trained on their artwork without consent.^[3] Getty Images also sued Stable Diffusion for copying more than 12 million images from its database without permission or payment.^[4]

We are still early in the development of AI so I expect the litigation to continue. But, despite these challenges, I suspect that artists will continue to use AI for its creative potential, even if the resulting work cannot be copyrighted. AI will be used as a tool to generate assets more than it will be used to generate full sequences or films, due to copyright laws.

Deep fakes, the dark side of technology

As we delve into the era of AI-driven visuals, the malicious act of seamlessly placing someone in a video, and engaging in actions they never committed, poses a tangible threat. X (formerly Twitter) recently witnessed a surge of explicit deep fakes featuring Taylor Swift, underscoring the urgency of addressing this issue, especially in the context of the impending 2024 election year.

“People believe what they see,” and with an estimated 500,000 video and voice deepfakes projected to flood social media in 2023, the stakes are higher than ever^[5]. Deepfake technology, once costing thousands is now available for a few dollars, adding another layer of concern. As the year and AI technology progresses, the potential for misinformation and manipulation grows exponentially.

OpenAI recognizes this threat and asserts its commitment to preventing “misinformation, hateful content, and bias” generated by Sora.^[6] Collaborating with Red Teamers, cybersecurity experts adept at ethical hacking, OpenAI takes proactive steps to assess vulnerabilities.^[6] This ongoing effort seeks to strike a balance between technological advancement and ethical responsibility in the evolving landscape of AI-generated content.

How to Adapt to AI

Given the burgeoning stage of AI technology, the perceived threat of its potential to disrupt or overhaul livelihoods is undeniably real. Its growing presence in the creative industry necessitates our adaptation to the inevitable changes it brings. While businesses may initially attempt to entrust AI with crafting their marketing campaigns, the intricate nature of controlling AI will become apparent, leading them back to the expertise of human artists.

Another theory postulates that AI could risk becoming mundane due to the emerging trend of training AI on existing AI, fostering a feedback loop.^[7] In contrast, human creativity remains inherently innovative. While AI serves as a valuable tool for inspiration, it tends to recycle ideas. On the other hand, artists possess the capacity to produce work that stretches the boundaries of what AI can achieve.

Addressing the legal front, there is a foundation of legal frameworks designed to safeguard artists, with ongoing improvements anticipated. The key lies in artists embracing AI as a complementary tool rather than a wholesale replacement for their skills. Artists and AI need to achieve a harmonious coexistence, preserving the unique and irreplaceable touch of human creativity in the face of advancing AI technology.

Learning AI and How to Prompt

I was listening to a video with an AI expert who explained that AI will not directly take your job. Someone who knows how to use AI will take your job. If you want to be a successful artist in the age of AI, learning how to use AI is imperative.

So, where can you learn how to craft a prompt that delivers the desired results? Several colleges and universities now have Generative AI boot camps, many offering online courses. Online training, for example, Coursera and LinkedIn Learning, also have classes and courses.

edX currently has over 300 courses on Artificial Intelligence at this time. edX: Prompt Engineering and Advanced ChatGPT This short course explains how prompts work in ChatGPT, and it is free if you do not need a certificate.

Prompt Engineering for ChatGPT from Vanderbilt University is about 18 hours of self-paced learning with very high reviews. The course is free if you do not need a certificate.

For an introduction to AI, Google: Google AI for Anyone is a 4-week course that takes you through how AI is used in the real world and explains concepts like Neural networks. The course covers what AI programming looks like. The course is free if you do not need a certificate.

Related articles

• The 10 Best Free Prompt Engineering Courses & Resources for ChatGPT, Midjourney & Co. by Maximilian Vogel.

References

1. Why This Award-Winning Piece of AI Art Can’t Be Copyrighted, Wired, by Matthew Allen, September 6, 2023.
2. New US copyright rules protect only AI art with ‘human authorship‘, The Art Newspaper, by Daniel Grant, May 4, 2023.
3. In a Blow for Artists, a Federal Judge Has Sided With Three A.I. Companies in a Copyright Dispute, artnet.com, by Adam Schrader, October 31, 2023.
4. Getty Images sues AI art generator Stable Diffusion in the US for copyright infringement, The Verge, by James Vincent, February 6, 2023.
5. Deepfaking it: America’s 2024 election collides with AI boom, Reuters, by Alexandra Ulmer and Anna Tong, May 30, 2023.
6. Safety, Sora AI, Openai.com
7. A.I. Art Experts Dismiss Fears That the Technology Will Destroy Humanity. But It Could Make Culture Really Boring, artnet.com, by Adam Schrader, June 21, 2023.