The development of “systems of knowing and understanding” provide universal languages researchers can employ to tackle the myriad challenges presented in the development of Artificial Intelligences.
In the video posted here, Andrew Ng offers a brief ontological examination of the keys issues surrounding Artificial Intelligence in his presentation, “The Future of Robotics and Artificial Intelligence” (Stanford University, STAN 2011).
Gruber (1992) argued that, “An ontology is an explicit specification of a conceptualization. The term is borrowed from philosophy, where an Ontology is a systematic account of Existence. For AI systems, what "exists" is that which can be represented. When the knowledge of a domain is represented in a declarative formalism, the set of objects that can be represented is called the universe of discourse” (Link).
In his presentation, Ng suggested that two main issues are paramount in the development of AI: Control and Perception. Control refers to a robot’s abilities to navigate and interact in physical spaces. Perception refers to a robot’s abilities to “see and understand” the world around it. The exposition of the narrative focused mainly on the dynamic issues related to robotic perceptions of environment, ability to contextualize information it gathers, and abilities to “read and understand” its social environments.
Ng draws from neuro-scientific approaches to argue for the development of foundational, simple algorithms that can drive robotic AI abilities to “perceive.” AI neural networks can be crafted, informed by and patterned after, human neuro-biological systems that control visual and auditory processing. This draws out an interesting area of inquiry that I will explore in future posts: namely, how does our “human” understanding of AI, our collective cultural technological competencies and cultural technological value structures, inform they we actually conceptualize and create AI?
In his 2006 presentation at the Human-Robot Interaction Seminar (Fachberriech Informatik Universitat Dortmund) titled, “Recognition and Simulation of Emotions,” Kleine-Cosack discussed the exigencies surrounding AI perception and emotion recognition. I will examine the issues of emotion perception in AI presented in his paper in future posts. In sum, he argued that “the acceptance of autonomous and especially humanoid robots will be directly dependent on their ability to recognize and simulate emotions.” An especially useful context to place Ng’s overview of AI cognitive perception. (Link).
AI abilities to place sensory information (perceptual information) in social context, and the development of unique AI responses to those sensory stimuli, are a whole other ballgame. Yet, the formalization of “systems of knowing” provided here take a large step toward integrating robots into human experiences.
Philosophical inquiry has demonstrated the traps inherent in developing formalized (ontological) systems and vocabularies, yet I am reminded of a phrase from my favorite science-fiction opus, Dune, when Thufir Hawat (one of the many characters in the Dune series that characterizes the dynamic range of transhumanistic expression) states, “the first step in avoiding a trap, is to know that the trap exists.”