Creative Artificially-Intelligent Agents for the Arts: An Interdisciplinary Science-and-Arts Approach

Coordinator:
Jonas Braasch

Advisors:
Selmer Bringsjord, USA
Ted Krueger, USA
Johannes Goebel, USA
Pauline Oliveros, USA

Artificial Intelligence (AI) has made impressive progress since its start in 1956. It now influences our daily lives, as AI systems are an integral part of consumer technology today, from SIRI to automobiles to Semantic Web. However, while AI systems can be very successful if they are precisely told what to do (e.g., perform a parallel parking task, play chess), they are usually useless if the objectives are not clearly spelled out. They can learn along a precisely given trajectory (e.g., to learn to understand spoken text or compose an instrumental music piece in the tradition of JS Bach), but they don’t break rules to produce something more exciting. Deep Blue can play chess, but if you present it with a game implemented on a chess board, it will be lost. In short, machines are simply not very creative. The idea of this white paper is to form an intellectual think tank to overcome existing roadblocks and investigate alternative strategies in AI. Among the items that will be discussed is the implementation of design oriented processes for AI systems. Artists and designers often work on a less hypothesisor goal‐driven approach as compared to scientists and engineers; they pursue an open‐ended, purely experimental approach instead, where the outcome of each phase informs the next one, not necessarily having a fixed goal in mind. Along with this approach, there is a need for better AI evaluation systems that can judge the outcome more freely than just examining the results along an externally given set of rules. Using the experience of artists with the abstract, can we make agents more creative by allowing them to be continuously evaluate what they accomplish? How can we create AI systems that can develop and evaluate their own concepts? Part of this discussion will include the creation of a network for more complex AI systems that simulate several areas of the brain, or the abstract AI equivalent, simultaneously, by using a meta‐concept to connect existing AI modules using a UDP protocol in a computer‐cluster network. Another central aspect are systems that can draw on different algorithms to perform a task, making the selection part of the creative process. Along the same lines, we can look into web data‐mining methods that allow these machines to receive information beyond what is given to them by the experimenter.

Action Items

1) Complex Systems with Modular Architecture and Interchangeable Data Format

Roadblock: A lot of specialized software exists to simulate certain aspects of intelligence from computational auditory scene analysis algorithms to logic prover. In general, it is still very difficult to combine these specialized systems to complex systems simulating multiple parts of the central nervous system.

Opportunity: Enable a dialog to find better ways to standardize communication protocols between different systems and to port algorithms to a unified platform for creative intelligent systems

Proposed Action: conference or symposium to start dialogue

Stakeholders: university-based groups, gaming and entertainment industry

2) Agents that can handle abstract media and techniques

Roadblock: In engineering and science related disciplines a common approach is to copy the human body in both form and functionality. Honda’s Asimo robot and Kaist’s Hubo are good examples for this approach. Sometimes abstract solutions provide a better functionality, for example robots from children and science fiction movies are often more sociable, but artists and designers often lack the technical expertise of engineers

Opportunity: bring both groups together to build on each others’ strength to build highly functional, powerful but abstract systems.

Proposed Action: conference or symposium to start dialogue

Stakeholders: university-based groups, gaming and entertainment industry

3) Need of creative synthetic characters that can develop new concepts

Roadblock: Over the last 40 years we have develop artificially intelligent agents that can produce creative work within a given context (e.g., compose music in the style of J.S. Bach), but system that go beyond this and develop their own concepts (e.g., a new composition style) do not exist yet (at least not in the sense that they can reflect and justify their actions).

Opportunity: bring together transdisciplinary groups of artists, psychologists, and engineers to elicit how humans complete these tasks and find ways to implement this knowledge to artificially intelligent systems.

Proposed Action: conference or symposium to start dialogue

Stakeholders: university-based groups, gaming and entertainment industry

Advertisements