Human–Artificial Intelligence Systems: How Human Survival First Principles Influence Machine Learning World Models

Abstract

World models is a construct that is used to represent internal models of the world. It is an important construct for human-artificial intelligence systems, because both natural and artificial agents can have world models. The term, natural agents, encompasses individual people and human organizations. Many human organizations apply artificial agents that include machine learning. In this paper, it is explained how human survival first principles of interactions between energy and entropy influence organization’s world models, and hence their implementations of machine learning. First, the world models construct is related to human organizations. This is done in terms of the construct’s origins in psychology theory-building during the 1930s through its applications in systems science during the 1970s to its recent applications in computational neuroscience. Second, it is explained how human survival first principles of interactions between energy and entropy influence organizational world models. Third, a practical example is provided of how survival first principles lead to opposing organizational world models. Fourth, it is explained how opposing organizational world models can constrain applications of machine learning. Overall, the paper highlights the influence of interactions between energy and entropy on organizations’ applications of machine learning. In doing so, profound challenges are revealed for human-artificial intelligence systems.

Safety of Human–Artificial Intelligence Systems: Applying Safety Science to Analyze Loopholes in Interactions between Human Organizations, Artificial Intelligence, and Individual People

“…Fundamentally, human intent is to try to survive. Accordingly, it can be expected that many individual people and human organizations will deploy technologies, including AI, to increase their capacity to survive [48]. This can involve human organizations exploiting loopholes even if this is done at the expense of many individual people [49,50].…”

Representing Human Ethical Requirements in Hybrid Machine Learning Models: Technical Opportunities and Fundamental Challenges

“…Instrumental inference involves inferring action options and consequences in the environment. Epistemic inference refers to updating internal representations in world models [14,49]. A machine learning world model, like all other world models, encompasses selected concepts and relationships between them to represent those aspects of the real-world system that it interacts with.…”

“…Furthermore, it is important to note that there are decades of evidence that organizations and individuals can continue to have opposing interpretations of exactly the same information, even if it is explained in detail with high-quality visual content [54]. Hence, as summarized in Figure 7, the representation of ethical requirements in machine learning world models needs to be carried out with consideration of human organizations' documented world models in their QMS, individual people's embodied psychomotor world models, and the potential for deeply rooted opposition between them [14,55].…”

“…Behavioral ethics can be summarized by the title of the paper: "Why good people sometimes do bad things" [9]. Human organizations that need their HML implementations to operate in alignment with human ethical requirements as defined in regulations, etc., need to take behavioral ethics into account [13,14].…”