A historical evaluation of artificial intelligence with focus on algorithms generating creative artwork

Abstract

The practice of history writing begins with the understanding and acknowledgment that the past cannot be completely reconstructed as it was. The trajectory of artificial intelligence (AI), the subjective agency of this dissertation, originates in the ancient efforts of humanity to emulate nature, and eventually intelligence on inanimate objects. This enduring fascination was driven by various motivations, primarily, by the desire to explain natural phenomena of the universe and the complex functioning of living organisms and the mind. The continual historical narrative of AI has been shaped by the mutual influence of diverse disciplines over time, propelled by strategic research and development during the Second World War, and eventually, AI was conceptually formalized as a field in the middle of the twentieth century. This dissertation aims to critically evaluate the existing historical accounts of artificial intelligence and compile them into a structured narrative, and subsequently uncover the origins of creativity within the context of AI. The mechanistic world view, promoted by the Scientific Revolution, significantly influenced the development of automatons and subsequent technological advancements. The history of AI is appreciated as continuation of this intellectual and technological narrative. The research in AI was inspired by arguments in philosophy and shaped by the conceptual discussions and formalizations in mathematics as a science. AI as an interdisciplinary field adapted theories from physics and was developed in interaction with psychology and linguistics. Eventually, the implementation of artificial intelligence was constrained by the technological capabilities and their inherent limitations prevalent at the time. Mathematics and physics were foundational in the establishment of artificial intelligence's theoretical framework. Development of formal systems, probability theory, concept of randomness, and arguments and proofs leading to the incompleteness of mathematics were influential. In physics, the theories of thermodynamics, primarily entropy, that influenced information theory, contributed to foundational thoughts about data and artificial intelligence. The concept of uncertainty, as established in physics, notably in quantum mechanics, anticipated a paradigm shift in artificial intelligence, moving from deterministic to probabilistic approaches. Moreover, the brain modeling theory that emerged in the early twentieth century, proposed the vision that neural networks were heterogeneous systems of neurons. During the mid-20th century, the concept of the intelligent machine gained prominence. The advent of the computer provided the essential platform for realizing the aspiration of implementing humanoid intelligence. Study shows that the story of artificial intelligence is an accumulation and assimilation of progressing events rather than a set of abrupt shift and sudden discontinuity over time. Nevertheless, the emergence of AI as a separate discipline is commonly related to the 1956 Dartmouth College conference event where the term artificial intelligence was officially used for the first time. The conference's proposal outlined specific problems to be addressed and key areas in the field of AI for future investigation. Since then, AI research and implementation technologies have progressed through multiple distinct phases. From the middle of the twentieth century, artificial intelligence was implemented based on symbolic manipulation, expert-systems databases, and data-driven machine learning approaches. The ability to process and utilize aesthetic elements differentiates human cognition from machine intelligence. Inspired by this assertion, building upon a comprehensive compilation of diverse AI historical narratives, this dissertation subsequently focused on the domain of artificial creativity, a topic that was central during the emergence of AI to the debate on whether machines can think. In this context, a conceptual exploration of creativity and its place within the field of artificial intelligence was performed. Creativity, a foundational feature of human intelligence, is widely related to aspects of chance, chaos, randomness, and unpredictability. The nascence of computer-generated arts coincides with the emergence of mainframe computers in the 1950s. Cybernetics and information aesthetics concepts influenced early art production experiments by the computer during 1960-70s. To discover whether computational creativity was implemented based on a common approach, early AI approaches across diverse creative domains of music composition, painting and poetry were examined. The emergence and development of computer creativity is evaluated within the broader context of AI history and in connection with the art movements. Contemporary approaches to artificial creativity are additionally examined to link historical accounts to future trajectories. Analysis across various creative domains revealed that early computer poetry, digital visual arts and music composition were implemented using similar underlying principles. In particular, a random generator was used in all fields to activate a few of the many interaction options within a system with defined boundaries, given inputs, and defined interaction rules, and as a result obtaining unpredictable but calculable outputs. The concepts of information aesthetics, inspired by cybernetics and information theory, were applied. Experiments were influenced by mathematical methods of stochastic and random algorithmic implementations. The computer was used as a laboratory environment to experiment with the formalized processes of innate human creativity. Researchers in each of the three domains selected for this study were exploring the creative processes and working on their algorithmic formalization. In the early experiments, rule-based algorithms were widely used. Pioneering research sought to elucidate the procedural mechanisms utilized by individuals during the creative process, rather than to produce outcomes of significant aesthetic or artistic value. Nevertheless, the endeavors contributed to the foundation of subsequent creative artificial intelligence models. This study was conducted following qualitative historical research. The research is based on textual analysis of published archival and contemporary literature and serves an interpretive approach. Critical interpretation of primary and secondary sources was performed. The research centered on the middle of the twentieth century, a pivotal period marked by the conceptual emergence of AI. Nevertheless, to maintain the integrity of the overarching historical account, foundational preceding concepts in the history of artificial intelligence are introduced in the report. The study concludes with an analysis of current state-of-the-art developments to establish a seamless historical narrative.