Heroic versus Collaborative AI for the Arts

Mark d'Inverno, Jon McCormackIJCAI 2015

Introduction

The prevailing narrative in Artificial Intelligence research often centers on "Heroic AI"—the pursuit of autonomous systems capable of matching or exceeding human performance in isolation. From Deep Blue's victory over Garry Kasparov to modern generative models, the benchmark for success has frequently been the machine's ability to act as a lone creator. However, in the realm of the arts, this model faces significant philosophical and practical challenges. This paper argues for a paradigm shift toward "Collaborative AI," where the goal is not to replace the human artist but to provoke and expand human creativity through sophisticated interaction.

The Problem with Heroic AI

In the arts, the "Heroic" model seeks to build an autonomous "computer artist" that functions without human intervention. The authors contend that this approach fundamentally misunderstands the nature of art. Unlike games with formal rules like chess, art is a social enterprise deeply rooted in human experience, culture, and communication.

The quest for total autonomy risks creating "creative lethargy" in humans, where automation leads to the homogenization of artistic output and a decline in personal skill development. Furthermore, the "Heroic" label is often a misnomer; in systems like Harold Cohen’s AARON, the true "hero" is the programmer whose artistic goals and edits are embedded within the software.

The SMART Agent Framework

To move beyond anthropomorphic descriptions of AI, the authors employ the SMART Agent Framework. This framework provides a technical hierarchy for distinguishing between different levels of system capability and intent.

The SMART Agent Framework defines a hierarchy from basic entities to autonomous agents with internal motivations.

The framework distinguishes between:

  • Agency: The ability of an entity to act on behalf of another to achieve a goal (e.g., a paintbrush serving an artist).
  • Autonomy: The capacity for an agent to generate its own goals through higher-level intrinsic motivations.

Collaborative AI systems are characterized by high agency but low autonomy. They are designed to interact, provide feedback, and "jam" with humans rather than setting their own independent artistic agendas.

Collaborative AI in Practice

A primary example of Collaborative AI is the Reflexive Looper. Unlike a standard recording device, this system "jams" with a musician by learning their patterns and playing back responsive variations. This interaction stimulates new creative directions that the human artist might not have explored in isolation.

Collaborative AI systems facilitate a "comping" style of interaction, where the machine responds to and challenges the human performer to find new creative routes.

By focusing on provocation rather than just efficiency, these systems serve as "difficult" partners. Instead of simply automating a task, they offer new creative routes based on a dynamic awareness of the artistic context.

Key Insights and Redefining Success

The shift toward collaborative models highlights several critical insights:

  • The Myth of the Lone Hero: AI lacks the social and biological "why" that drives human expression. Its "creativity" is often a pastiche of human experience rather than genuine expression.
  • Provocation as a Feature: Success should be measured by how much the AI helps a human artist reach new levels of expression, rather than whether it can "fool" an audience into thinking a human made the work.
  • Process over Product: Audiences are often more interested in the human struggle and process behind a work of art. AI that reveals the nature of the artistic process through interaction is often more compelling than a "black box" output.

Conclusion and Future Directions

The future of AI in the arts lies in participatory design that prioritizes the needs of artists. While challenges remain—such as designing systems that provoke without disrupting the "flow" state—the move toward "communities of agents" offers a promising path. By treating AI as a sophisticated partner rather than a replacement, we can ensure that technology enriches rather than diminishes the human creative spirit.

Summary updated

BibTeX
@inproceedings{dinverno2015heroic,
  author = {Mark d'Inverno and Jon McCormack},
  title = {{Heroic versus Collaborative AI for the Arts}},
  booktitle = {International Joint Conference on Artificial Intelligence},
  year = {2015},
}
Figure