Recent studies on Large Language Models (LLMs) and Large Reasoning Models (LRMs) reveal how AIs currently “think”. The insights underscore how elusive the nature of concepts like “meaning” and “understanding” are. Though LLMs excel at generating fluent, human-like language and providing explainable narratives, they are viewed as word predictors - but they show the capacity to cluster concepts in token space. LRMs, designed specifically for structured logical reasoning and multi-step problem-solving, demonstrate an ability to explicitly model reasoning but fall considerably short in complex tasks. Notables like Rich Sutton point towards deep reinforcement learning to deliver experience-based models built upon curiosity and causality, while multi-modal AIs today consume a range of information types including audio and visual. Embodiment and direct access to the physical world are next, as AI are connected to real-time sensors, incorporate world foundation models, and transition into robots. Clearly new kinds of minds are in the making - but how will they think?

Dr. Loki Jörgenson is a dynamic tech leader bringing 20+ years of private industry entrepreneurship and 7 years of academia to his role at Circle Innovation. As founder and CTO of Vancouver and Bay area tech startups, Loki provided strategic vision and direction, developing machine learning technologies for IT, fintech, and medtech. His PhD in computational physics (McGill) enabled him to commercialize data-intensive, ML and predictive solutions for B2B and B2C markets. Loki has been an active member of​ several research NCEs and has served as an adjunct professor of mathematics.