AI for Creative Arts, and a Data Culture to Propel It
- Share via
AI can be creative, and if you still believe it lacks a crucial human spark – how long before that spark is replicated and surpassed? While the neurological mechanics of AI creativity are inspired by that of human neurology, the details and raw materials involved differ.
At the end of the day, AI is a prediction machine built on data, models and computing infrastructure – though I, for one, have yet to be convinced that human intelligence isn’t as well. But that misses the point, and it certainly isn’t a reason to think AI doesn’t or won’t exceed human creativity. Alas, creativity takes many forms, and the most transformative innovations draw from multiple domains.
One enclave of creativity is the creative arts – visual, performance, music, design, and language. As entertainers, writers, artists, musicians, dancers and even chefs are thrust into a brave new world with AI, why and how must data culture evolve to transform this domain? Today, we witness the ability to interact with data in plain language at scale. This paradigm-shifting opportunity enables creative artists to extract massive amounts of inspiration essential to their craft.
It probably doesn’t surprise you to hear that the quality of AI creativity is proportional to the inspiration it draws from. And in AI, that inspiration is captured and stored in data. How must data culture evolve to manifest the quantum leap in creative arts that is being felt elsewhere from AI?
Data culture is defined as an organization’s shared beliefs, practices and values for using data to drive decisions, operations and outputs. Prior to the recent AI revolution, aside from some digitization, creative arts have largely been excluded from the collective consciousness around how we capture and store data for operating systems generating creative artistic output.
Artistic outputs can be measured and recorded through various modalities in decomposition, such as audio sound waves, written language and video imagery. A measurable audience and the impact of the art’s exposure or consumption can be input to generate new artistic outputs. The essential foundations: data, models, and computation constitute the basis from which applications draw. Diversifying the modalities of inputs and complexity, along with expanding the context and think-time, are unlocking the unthinkable in artistic AI creativity. AI powered systems now produce novel, exceptional, and precise artistic outputs, and this evolution demands a data culture that comprehends and considers decisions, operations, inputs and outputs of artistic modalities and the teams responsible for bringing ideas to life.
In a recent article I wrote on ontology for data, I talked about the importance of organizing data in a way that allows AI to extract the right information for training, and access the right information for inference. Like most things, the raw materials and how you use them makes all the difference. Data is no exception, and is the leading candidate for being the distinctive and fortified defense for a business – especially as AI applications continue to improve at making application development easy and fast at a low cost.
A well-designed ontology organizes knowledge in data to maximize the information extracted from it. This principle applies across various AI systems, including the training of Large Language Models (LLMs). LLM inference, for example, relies on tokenization, breaking language into units. This process serves as data input and output for LLM training and inference, illustrating one familiar example of how data organization is crucial for AI models.
Good principals for organizing and relating knowledge in its various modes and forms enable smart agent configuration and optimal collaboration among multiple agents and how they work together. Consider the example of producing an advertisement, a process at the intersection of big ideas, creative art, and distribution to audiences, complete with clear business objectives.
Traditionally, a team assembles images or videos, and the ad copy, in various forms for placement in specific media. This is slow and labor intensive and produces only a few variations of the ad. A creative system engineer, however, can prompt an AI system to develop an application. The system breaks up jobs for the collective of specialized AI agents tailored to build an application that produces virtually endless variations of an advertisement in endless formats. The system can also categorize the ads and their components to track performance and reuse their elements while experimenting with completely new ones. This dramatically cuts labor costs, speeds up turnaround and reduces friction between traditionally separate creative, media, and analysis teams and ultimately improves performance for the audience.
A data culture for creative arts is essential for transforming a creative arts organization or business for AI. Creativity, art, and the business of it require data in an AI inspiring form and a collaborative operational data culture for the creative engineers and artists building it.
