Foundations of Synthetic Consciousness
For the past few years I’ve been doing part time research in the realm of cognitive sciences, with emphasis on consciousness and emergent intelligence. From my findings, it appears that [artificial] intelligence can only emanate from [artificial] life, through the bottom up integration of biological and cognitive systems, with consciousness at its apex. In nature, consciousness seems to have arisen to facilitate self-preservation of complex organisms in dynamic environments. As such, my premise of successfully realising a complex and autonomous artificial intelligence relies on the successful creation of a fairly complex artificial organism (life), inclusive of a [synthetic/artificial] consciousness.
Consciousness is an extremely complex system, if we can call it as such, and one can only assume that it arises from the connectivity of hierarchical “components”. In its simplest form, we can hypothesise that the following components and systems form the basic elements from which consciousness arises:
- Somatosensory System: Allows organism to sense environment, provides awareness, and objectification of self within environment. Considering the sense of touch for instance: input levels impact homeostasis. Soft and light input may result in feelings of pleasure whereas high intensity input may result in feeling of pain, thus impacting behavior based on homeostatic rules in procedural system.
- Procedural System: Basic and elemental knowledge encoded in organism (dispositions). Provides foundations of self-preservation. Integrates concept of homeostasis. Purpose of the homeostatic system is to maximise wellness, ensure subsistence/existence, maximise one’s positive state of being. In other words, homeostasis serves to preserve the Self.
- Cognitive Reconstruction System: Provides means for representing events and states in an easily accessible and efficiently processed form. Creates internal images of internal states, external environment, current thoughts (including memory recalls), and intents (intent to take an action, move, etc).
- Memory Storage and Recall: Intelligence potentially arises from learning and experience, which can only be achieved with memory and recall mechanisms. Events must be recorded in such a way that they can be later retrieved in relation to current contexts and can be cross-utilized for different needs. Includes subsystems for Episodic and Semantic memory. World knowledge, memory and experience are key to semantic interpretation. One can only truly understand the meaning of “I’m eating a fruit” if one knows what eating is and what a fruit is. Furthermore, the proper and correct understanding of the meaning of an action, such as eating for instance, can only be achieved through experience. “The action that you are performing now is called ‘eating’”, “What you are eating now is a fruit”. From this experience, one can associate characteristics, context and feelings to words.
In addition to the core research, I’ve been experimenting with the above mentioned components and have achieved some interesting results. By developing elements of the homeostatic system, Episodic and Semantic memory, along with the integration of world knowledge, I was able to develop a system that is able to achieve a basic level of natural language understanding (NLU) without relying on Machine Learning or other statistical methods. In addition, the system began to display the concept of choice by selecting one item over another based on personal preferences, emerging from the codified psychological profile and personal experience.
Based on the promising early results, I intend on continuing reviewing the latest findings in Cognitive/Neuroscience research and improve my overall software architecture to facilitate the development of new and revised systems and components. I believe that true AI will only emerge from the development of Synthetic Consciousness and hope to contribute further in the field.