ASCII×GPT

Client

Internal

Exploration

Project Type

Code

Overview

We created ASCII animation art using tools that we built with AI.

Balancing Technology and Artistry

Safeguarding human creativity in the face of AI’s rapid advancement presents a complex challenge. This dilemma is not new to technological advancements, like the printing press or the camera, but the speed and scale of its impact is unprecedented. It has been a challenge to keep up and our posture is still one of curiosity.

We believe that understanding its benefits and risks to be vital to our commitment to valuing people.

While AI offers opportunities for improving efficiency, there are categories of machine learning that pose a potential threat to the intrinsic value of human creativity and the well-being of individuals in creative professions.

Starting with an open mind, let's consider the role of machine learning in daily activities like autocorrect, email spam filters, recommendations from streaming services, smart home devices, real-time navigation, and healthcare diagnostics.

These applications appear to be ethically acceptable and could provide a framework for considering AI's role in creative pipelines. The challenge lies in balancing the exciting potential for increased productivity with the threat of devaluing the unique beauty of handcrafted artistry.

That brings us to the question at the core of this exploration: Can AI complement human creativity without overshadowing it?

Building Tools with AI

Our approach was to steer clear of image generation tools and instead utilize LLMs (Large Language Models). Could we create an animation tool with the help of AI to draft code. We focused on ASCII in HTML5 because of its relation to language and coding patterns (something LLMs are good at). The initial explorations showed promise, but as we integrated more complex features, the increase in code errors demanded significant human intervention. We also grappled with accuracy and memory drift issues, leading to a strategic division of the codebase into smaller segments for targeted training.

After investing over 250,000 tokens and dedicating several weeks of refinement and troubleshooting, the tool began to work as intended. What we have found is just how important it is to define AI's contribution to any project: understanding the boundary between its ability and the distinct nature of human artistic expression. Our tool effectively utilized AI's computational power, yet the final creative outputs were deeply reliant on taste, decision-making, and problem-solving skills that only humans can provide.

This exploration has cast some more light on both the limitations and the capabilities of AI as it relates to creative pipelines. As we navigate the merging paths of man and machine, it's vital that we use these advancements thoughtfully, safeguarding the domain of human artistic expression.