Digits that are not: Generating new types through deep neural nets
This addresses the barrier in creativity research for AI systems to develop their own value for novelty, though it appears incremental as it builds on existing deep learning methods.
The paper tackles the problem of artificial creative agents needing external value functions for novelty by proposing a knowledge-driven creativity approach that allows systems to explore based on learned referential objects, resulting in the generation of coherent sets of new digit types not belonging to known categories using a deep generative autoencoder.
For an artificial creative agent, an essential driver of the search for novelty is a value function which is often provided by the system designer or users. We argue that an important barrier for progress in creativity research is the inability of these systems to develop their own notion of value for novelty. We propose a notion of knowledge-driven creativity that circumvent the need for an externally imposed value function, allowing the system to explore based on what it has learned from a set of referential objects. The concept is illustrated by a specific knowledge model provided by a deep generative autoencoder. Using the described system, we train a knowledge model on a set of digit images and we use the same model to build coherent sets of new digits that do not belong to known digit types.