Aegis Location Bible (Retrospective)
A review of AI in location bibles, assessing its artistic, ethical, and legal effects.
To all of you accompanying me on this deep dive into the artistry, ethics, and legality of integrating AI into the realm of cinema, I offer my heartfelt thanks. While the outset was marked by doubt and I wasn't spared critique, the tide seems to be turning. As we venture further, an increasing number of individuals are not just watching from the sidelines but actively joining the conversation, a truly noteworthy shift. From a humble following of 10 at the start, we've burgeoned to many thousands of daily readers. The journey is still in its infancy, but a vibrant community centered around Shepard’s Tone and our ambitious AI venture is unmistakably taking shape. It's remarkable to see this community being primarily fueled by screenplay writers, producers, directors, and specialists in animation and conceptual art – all of whom share a fervent love for storytelling and cinema. Your participation is both a testament to the relevance of this venture and a source of encouragement. A sincere thank you for being on this journey with me.
Now on to the important stuff - have I lost my soul yet?
The journey of utilizing AI to navigate the filmmaking process from script to screen has been a profound one for me, particularly in the realm of ethics. Initially, my ethical stance on AI might best be described as naive. As I delved deeper into its practical application, it became evident that forming a comprehensive ethical perspective would demand persistent effort and immersive engagement. Through hands-on experience, I discovered a dynamic shift in my ethical positions concerning AI. Some apprehensions I held were dispelled, while in other areas, I realized I had underestimated potential concerns. Interestingly, certain anticipated uses of AI proved to be misconceptions, and other presumptions I had were debunked. This intricate journey has reshaped my understanding, and I now find myself armed with a fresh set of preconceptions and insights.
My false preconceptions
When I started this process, I assumed that the predominant utility of AI would be to reference and amalgamate characters and locations from existing films. This raised a pressing ethical question: Was I inadvertently committing intellectual theft? However, when I engaged in the tangible process of producing character and location bibles, my initial notions were challenged. Merging elements from established artworks didn’t resonate well with my artistic sensibilities, and equally as important, nor did it yield the high-quality outcomes I anticipated. I had also speculated that I'd be frequently directing ChatGPT to "write in the style of…” specific authors or filmmakers, thereby revisiting the ethical dilemma of theft. Yet, in reality, such attempts led to results that felt profoundly inauthentic and lacked the requisite quality for professional production. If one were to enumerate the potential misuses of AI that border on intellectual appropriation, I’d confidently argue that each such application would render outputs that seem contrived and are glaringly subpar in production value.
My current intuition suggests that professionals striving to employ AI in filmmaking will naturally drift away from methodologies that mimic or replicate existing works. This inclination is likely fueled by an innate commitment to the project's quality and the industry's aversion to palpable replications.
A THEME TO REMEMBER: In professional filmmaking, AI emerges as a tool of augmentation, serving as research and writing assistance, complementing the originality and innovation of the writer's foundational ideas and artistic vision. The true essence of the art remains firmly rooted in human creativity. Contrastingly, in the realm of social media, there's a propensity to lean towards the lowest common denominator, with borrowing, and at times outright replication, becoming prevalent. Here, AI's potential might be channeled in ways that prioritize virality and trends over originality and authenticity.
My biggest concerns after this exercise
My foremost apprehensions surrounding the use of AI, especially in crafting location and character bibles for screenplays, revolve around the potential inaccuracies and biases embedded within the technology. If we lean on AI to ensure that our characters, locations, and the worlds we sculpt are historically accurate, then we are inextricably tethered to the AI's own interpretation of history. The unsettling question arises: what if external forces, be it foreign governments, political factions, or religious groups, find ways to influence or manipulate AI's understanding of "history"? There's an imminent danger of obfuscating, or even erasing, the truth of our past. Furthermore, AI's inherent biases can cast a shadow on its output. For instance, when I probed ChatGPT about the disciples for the Seraphim or the leader of the Fawda, it consistently chose male figures, revealing a gender bias. More alarmingly, when I sought insights about historically Arabic characters, I encountered inherent biases in ChatGPT. Some were relatively harmless cultural assumptions, like food preferences, but others bordered on racial prejudice. Such biases can inadvertently perpetuate stereotypes and misconceptions, distorting our representation of characters and locations. This raises profound ethical concerns about the credibility and inclusivity of AI-generated content.
Oddly, this makes me happy because it lets me explore the dangers of AI in a more personal way through telling the story of the Shepard’s Tone. The core themes of Shepard’s Tone revolve not only around the dichotomy of ORDER versus CHAOS but also dive deep into the narrative of artificial intelligence—how it once led to humanity's downfall and is presently re-emerging as a looming existential threat. By employing AI to elucidate the ways it could potentially challenge humanity, I gain a multi-layered understanding, enriching the depth and precision of my story. This dual benefit not only crafts a more compelling and authentic narrative but also refines my ethical perspective regarding AI—and hopefully, that of others. Illuminating the ethical quandaries of AI within a film centered on AI instigates a meta-discourse, urging the audience to critically engage and reflect. This approach fosters a healthy and constructive dialogue, allowing for a comprehensive exploration of the myriad ethical implications of AI. Such introspection is vital as we tread forward in this AI-driven age, ensuring that art doesn't just mirror life, but also guides and informs it.
Using ChatGPT: What works and what doesn’t work?
As I journeyed through the intricacies of employing ChatGPT in creating the location and character bibles for Shepard’s Tone, I discerned a pattern. At its core, AI, particularly ChatGPT, functions akin to an INPUT/OUTPUT machine, with distinctive features that, when understood and maneuvered adeptly, can produce optimized outputs tailored to the user's intent. However, there are certain operational dynamics one must grasp to make the most of this tool.
A salient realization during my experimentations was the existence of a "RETURN curve". When inputs are minimalistic, like posing a request as broad as “write me a screenplay about…”, and expecting a comprehensive output such as an entire screenplay, the results are suboptimal. The expectation is grandiose compared to the provided direction. Conversely, when I present maximal input, offering an exhaustive description of, say, a location, and seek a minimal output—like restructuring my words logically—the AI excels, acting more as an efficient editor.
Nestled between these extremes is a Golden Equilibrium. This entails delivering detailed input on a specific component of the screenplay, and then seeking a defined expansion from ChatGPT, bounded by specific parameters and guided by precise instructions.
Yes, that does sound complex. But, let's delve deeper to clarify.
The Golden Pattern
Let's use the creation of a location for Aegis as an example. The initial step involves outlining a hierarchy of key locations within this meta-location. For this instance, we’ll zone in on the Luminara Spire. I then segment the Luminara Spire into its 15 hierarchical components—each floor dedicated to one of the disciples of the Seraphim, one for the Shepard’s Tone Bell, the Grand Entrance, and another for the Crypt.
It's crucial to note that all this segmentation is my handiwork, not ChatGPT's.
With this structure in place, I established a pattern, primarily focusing on the RETURN output related to just ONE floor of the bible. The process to engage ChatGPT consists of distinct steps:
Step One - Preloading Constraint Data: I initiate a new chat and pre-load it with extant information pertinent to the forthcoming prompts. As an illustrative example for one floor of the Luminara Spire, I upload details like the history of Aegis, an account of the Great War, the chronicles of the Seraphim, the distinct character sketches of the disciples, and the overarching bible for the spire. My instruction to ChatGPT is simple: "read this..." followed by pasting the relevant descriptions and then hitting RETURN.
Usually ChatGPT politely let’s me know that it’s read the information… sometime it just starts writing something - ignore this for now.
Step Two - Initializing a Properly Constructed Prompt: My prompt to ChatGPT is multifaceted. All of the following is within one prompt.
Directive and Description. Component One encompasses a broad description I offer for that particular floor.
Postulations and Expansions. I ask ChatGPT to POSTULATE and EXPAND on the terse description I've provided. This expansion is DIRECTED towards generating details like the ambient smells, objects populating the space, and the emotional resonance of the locale.
Constraints. I instruct ChatGPT to ensure all postulated elements harmonize with the WORLD RULES delineated in the pre-loaded content, thus ensuring consistency within the established universe.
Negative Constraints. This assists in maintaining a distinctive feel for each floor, urging ChatGPT to steer clear of replicating objects, aromas, or emotions present in other parts of the spire.
It's important to realize this becomes an iterative process, especially as I advance in crafting the narrative of each floor. With everything set, I hit ENTER, eagerly awaiting ChatGPT's crafted output.
EXAMPLE PROMPT: In a neutral and balanced tone, write me a location bible, in paragraph form, for the second floor of the Luminara Spire. Note that the second floor is the domain of Caelum - Keeper of Astronomy: Lithe, silver-eyed with constellation tattoos. Strength: Predicting celestial events; Weakness: Fear of darkness. The floor is a fascinating workspace and study for Caelum’s craft. EXPAND on this location bible definition by POSTULATING how one feels when entering the space, what objects does one see, what smells exist, what is the layout, what color persist. Tell me the origin story of how this room was created and what secrets lie within this floor that Caelum keeps to himself. CONSTRAIN the POSTULATIONS by the previous paragraphs in this chat. And DO NOT have any POSTULATIONS that overlap with other floors in the SPIRE.
Step Three: Infolding - Iterating and Refining. The post-generation phase is paramount, emphasizing iteration and refinement. But there's a nuanced art to this step. Let's differentiate between the optimal approach and a pitfall-laden one. Suppose upon perusal of ChatGPT's output, you desire the inclusion of a covert workspace brimming with prohibited tech.
THE WRONG WAY: Submitting a fresh prompt like “can you add a secret workroom with banned technology to the location bible” is a recipe for chaos. This approach prompts ChatGPT to overhaul the entire bible, birthing entirely new data. More alarmingly, vital fragments of the original directive, in terms of tokens, begin to fade. Engage in this iterative pattern for a mere 10 cycles and behold the disintegration of constraints, postulations, and the world order. The AI's grasp wanes, and you're essentially flying blind.
THE RIGHT WAY: Navigate to the initial prompt and utilize the adjacent edit option. Inject your additional details here. While ChatGPT will still regenerate the content, this method ensures fidelity to the pre-established world order, constraints, and directives.
But what if elements of the previous iterations appeal to me?
Two pathways exist. First, archive the original output externally, perhaps in a Google Doc, rendering it easily accessible. When instructing ChatGPT to fashion a new narrative, this archived version can later be juxtaposed with the new, guiding ChatGPT to meld both seamlessly. Whether you dictate specific elements to retain or grant ChatGPT autonomy, the choice is yours. Alternatively, post-output, append a prompt specifying retention of all extant details, merely supplementing with the additions. This engenders a self-contained constraint mechanism, but heed caution. This mode does erode ChatGPT's overarching understanding of the world order. It excels for final touches but demands vigilant oversight to ensure consistency with the broader narrative landscape.
What I’ve found is that about 3-5 infolds are sufficient to produce stunning results.
Step Four. Recursive-Infolding. Recursion is a principle wherein a function, process, or algorithm achieves an outcome by repeatedly applying itself within its own structure. A classic example is the mathematical factorization, where a number is broken down by repeatedly dividing it by prime numbers until only primes remain. Introducing the AI concept of "Recursive Infolding" extends this notion. Recursive Infolding deals with continuously applying constraint mechanisms within a defined world order, ensuring congruence and consistency throughout the narrative's hierarchy.
Imagine constructing location bibles for individual floors of a building. Once all floors are detailed, the broader setting — the entire building, the district, the world, and even the universe — is established. However, as the scope expands, maintaining consistency becomes increasingly complex. To ensure the absence of conflicting details, such as overly proximate object descriptions or replicated secret workshops across floors, one would traditionally need to comb through the entirety of the content meticulously.
Enter AI.
By initiating a new CHAT and pre-loading all floor bibles into it, ChatGPT can be directed to undertake an exhaustive review. The prompt instructs the AI to reconstruct all the floor bibles, retaining unique details while identifying and modifying any overlapping or incongruent elements. By reevaluating the entire structure and applying the constraints recursively, the AI effectively 'infolds' the narrative, preserving its integrity.
Conclusion.
Here's my revelations I've concluded in my exploration with ChatGPT: instead of diminishing the role of a writer, as I assumed AI would, a proper use of AI actually prompts more profound writing. Locating the Golden Equilibrium pushes writers to actively engage with their craft, ensuring the preservation of the art of storytelling, even while leveraging technology. As I've seen ChatGPT magnify and embroider upon my foundational ideas, I've unearthed historical insights, unraveled nuances in objects, and encountered intricate interconnections that birthed fresh creative avenues.
But there's a caveat; collaborating with AI requires an entirely new skill set - one that I’m not sure established writers would necessarily embrace. This “prompting” skill set demands exploration, learning, and eventual mastery, and we've merely scraped the surface. Those who adeptly harness this potential might soon rival the prowess of entire writing teams, offering an output that previously took droves to achieve. Envision a future where artistry flourishes, not despite, but because of AI. We could produce incredibly detailed character and location bibles, as well as cohesive world orders. This meticulous craftsmanship could render our characters and their universes palpably genuine, deepening engagement and elevating the essence of storytelling.
But I guess it’s all a matter of each person’s perspective isn’t it.
Next week, we see our first pictures of characters and locations… mmmmmm.
What is The Brief and Who should read it?
I release a weekly digest every Friday, tailored for professionals ranging from executives to writers, directors, cinematographers, editors, and anyone actively involved in the film and television domain. This briefing offers a comprehensive yet accessible perspective on the convergence of technology and its implications for the movie and TV industry. It serves as an efficient gateway to understanding the nexus between Hollywood and Silicon Valley.
Who am I?
I'm Steve Newcomb. Functionally, I’m a recovering Silicon Valley founder that is finally old enough to have a bit of care. I’m perhaps most recognized for founding Powerset— it was the largest AI and machine learning project in the world when I founded it. It was later acquired by Microsoft and transformed into something you might recognize today - Microsoft Bing. Beyond Bing, I had the privilege of being on the pioneering team that witnessed the inaugural email sent via a mobile device. My journey also led me to SRI (Stanford Research Institute), where we laid the groundwork for contemporary speech recognition technology. Additionally, I was a co-founder of the debut company to introduce a 3D physics engine in Javascript. I've held positions on the board of directors and contributed funding to massive open source initiatives like NodeJS and even the largest such project, jQuery. My experience extends to academia, having been a senior fellow at the University of California, Berkeley's engineering and business faculties. Recently, I ventured into Layer 2 internet protocols and assisted a company named Matter Labs in securing $440 million in funding to bolster their endeavors.
What am I doing besides writing these posts?
Typically, I allocate a year between groundbreaking ventures. My exploration for the upcoming project commenced in May 2023, and the sole certainty is its nexus with the film, television, SMURF, and AI domains. Sharing insights on my research endeavors helps me discern between feasible prospects and mere illusions. My hope is that for this venture, I appropriately consider the ethical and sociological repercussions.
If you are interested in contacting me, being interviewed, being helped, or yelling at me, my email is steve.e.newcomb@gmail.com.