Exploring the "Upside-Down" Creative Act with AI in Art

In this new era of Generative AI, the creation of art, the role of intent, and the actions taken to produce visual images are undergoing a transformation. I can’t yet call this a core evolution to what it means to make art, but it’s definitely a fascinating new paradigm that warrants exploration.

Traditional Art Creation: The (Oversimplified) Step-by-Step Approach

📈 Linear Progression: Traditional artists start with a blank canvas, building their work through deliberate, sequential actions.

✍️ Development of Style: Style emerges as a byproduct of the creative process, tied to the resources, sequence of actions, and the artist's personal touch. To replicate a specific style, one must reproduce the series of actions that led to its creation.

💪 Consistency Through Actions: The process is consistent because it's grounded in the artist's hands-on practice and refinement. The materials and processes available to them.

AI-Assisted Art Creation: Starting with Wide Possibilities and Finding your Narrow Path

💡 Diverse Outputs from the Start: When using AI, artists receive varied, highly executed outputs right from the beginning, thanks to well-trained models and well crafted prompts (not to mention LLMs that can improve prompts behind the scene in guided ways).

🎨 Unconstrained Style Application: AI models can “easily” represent a concept across a wide range of styles, providing an unconstrained approach to general stylistic variation. Without the costs of production as an obstacle, the new challenge then lies in refining and curating these outputs to match the artist's vision.

🔧 Curating & Refining the Vision: The artist’s role shifts to selecting and refining AI outputs to achieve the intended aesthetic. These outputs can then be re-fed to the model via training to further strengthen it’s knowledge of your style or subject. Traditional creative skills like photography, sketching, animation, photobashing, etc all drive tremendous value in this stage. Model finetuning and Prompt engineering also carry serious weight as needed skills in this phase.

Implications for the Creative Production Market

📈 Shift in Creative Development: Traditionally, style is chosen early in the campaign development, driving the entire creative process and budgeting. With AI, the marginal cost of applying different styles is reduced, shifting the focus from selection to curation.

💵 Reduced Marginal Cost: The ability to experiment with multiple styles early in the process without committing significant resources allows for more exploration in outputs. (good or bad thing? depends.)

🛠️ New Skill Sets Required: Ensuring consistency now involves curating and maintaining an approach to style rather than building it from the ground up. Prompt engineering, Fine-tuning AI models, and workflow development are key parts of this process.

Visual Diagram: [A diagram contrasting traditional style development with AI-assisted style curation, showing how the focus shifts from foundational development to curating and maintaining consistency.]

To Sum It Up:

Traditional Artistic Development Model:

An idea is developed upward through phases of work, refinement, and exploration to yield a final work with a resulting style.

AI-Assisted Artistic Development Model:

An idea generates an infinite array of potential “final outputs” immediately. However, these are not natively grounded by any conceptual knowledge of WHY it is producing what it is producing. This underscores the importance of the WHY factor in the process. The creative “work” here involves understanding the model's biases, providing it appropriate additional data to comprehend your task, refining its outputs, and making it an enjoyable and productive collaborator for your vision.

AI-assisted artistic development can dramatically uplift the value of the humans on your team, but it requires some design-thinking and collaboration with the models via finetuning and workflow design.

Ultimately the best performing creative teams will harness what’s best out of both of these paradigms. It’s not an either/or scenario!