The rapidity with which AI has become part of daily life for wide segments of society raises many concerns. Not many users, for example, are trained to manage their relationship with the machine and its pseudo-intelligence, and it is not uncommon for people to confuse machine responses as those of an intelligent, even emotional, agent. Partly, this is a matter of education in understanding of what exactly the machine (aka intelligence) is, compared to their own human nature. It is a hole that few corporations and providers of AI are even trying to mitigate, and this is no small matter. In this talk we shall compare the basic ‘architecture’ of rational human decision making compared to the way in which LLM’s reach their conclusions, and how agents, our alter-egos in the machine and on the internet, are essentially (or should be) extensions of ourselves into the cybernetic world. As humans and machine become more integrated, there is a real risk of humans forgetting the central purpose of their creation. Additionally, where humans see uncertainty, the machine often provides answers with unwarranted confidence. Without injecting uncertainty and doubt in machine responses, and cross-checking, erroneous myths and misguidance could lead to disaster. Through tests, scifi, myths and tales, technical analysis, and even illustrative art, we’ll plumb some of the depths of the human/machine combination to better comprehend the abilities of the machine, its current limits, dangers, and how we may manage that relationship – and why it needs managing in the first place – in order to achieve what the author considers is the primary goal: the benefit of humans. For some, however, the human-machine world or just the machine is an alternative and conflicting goal. Which of course raises the all-important initial question that is so needed in decision-making: what is the goal, and do we really want it?
Dr. Eddie Robins has a Ph.D. and Masters in Physics and Electrical Engineering from the University of Manchester, and an undergraduate degree in Physics from Imperial College, London. His career has spanned forty plus years over a range of industries and disciplines. From 1981 to 1989 he developed advanced algorithms and methods for data analysis, data analytics and data management in Canada’s National Fusion facility based in Quebec, Canada. In 1995, he joined Arlington Software in Montreal as VP and Chief Scientist to provide the underpinning algorithms and methods for multi-criteria decision-making products. These were applied to complex, large-scale decision problems in industry and governments in Canada, the US, and the UK. These decisions often required transparency to ensure liability risks were minimized in the decisions. Dr. Robins served as both consultant and lead in many projects which ranged from government and corporate policy decisions, to decisions involving complex, technological analysis with multiple cross-disciplinary teams. Though the decision process and algorithms may be thought of as elements of AI, they differ in several important aspects, the most major of which is keeping humans centered on the process, and the goal or goals focused on human needs. The machine, the algorithms, the people and the decision leader / mentor were necessary components to reach consensus (Pareto optimal.) At EMC (now part of DELL), Dr. Robins lead several corporate wide efforts to manage risk and corporate and customer exposure to loss, largely through models and data analytics. Here he employed a variety of mathematical techniques, including Bayesian and neural nets, fuzzy logic, device physics, and Markov models. Access to large data sources, modeling human behavior as well as machine and component characteristics were necessary attributes, which made him estimate failure in the light of the machine-human link. This gave him a unique perspective of large-scale human-machine interactions, and the potential for corporate and personal liability. From this experience, he sees the need for guidance and guidelines in the use of such complex systems, and to more fully understand where human and machine can best find a working relationship for the benefit of (hopefully) the human world.
