Conference Presentations
Inverted cognition: Toward minds that begin with output and derive goals retroactively
Lee R. X. AGI-25, 1(18), 362-374. Aug. 13th, 2025 | Reykjavík, Iceland. The Artificial General Intelligence Society.
CephNeuroAI Initiative, Division of Convergence Research, MIT Sigma Xi.
Summary | This proceedings paper proposes a model of inverted cognition in which intelligent agents act first and derive goals retroactively, challenging the dominant view that behavior must be guided by pre-specified intentions. Drawing on cephalopod neuroscience, predictive processing, and cybernetics, it shows how goal-like behavior can emerge from exploratory actions and be consolidated through meta-cognitive interpretation. The resulting algorithmic framework offers a flexible, corrigible approach to artificial general intelligence (AGI), where goals are not imposed but continuously discovered and revised through interaction.
Award | 2025 Kurzweil Prize — “for offering a provocative rethinking of how artificial minds might be structured”
Decentralized decision-making and the foundations of agency
Lee R. X. SPAN 2025, 16. May 3rd, 2025 | Saint Louis, MO, USA. The Society for Philosophy and Neuroscience.
CephNeuroAI Initiative, Division of Convergence Research, MIT Sigma Xi.
Summary | This talk examines whether systems like the octopus—whose arms exhibit autonomous sensing and action—can be considered unified agents despite lacking centralized control. Drawing on empirical findings and the framework of minimal agency, it argues that coherence toward shared goals emerges through decentralized communication, with “goals” functioning as explanatory constructs attributed by observers rather than internally represented drivers. This analysis reframes agency as a relational and context-sensitive property, challenging assumptions about intentional action and informing how we understand intelligence in both biological and artificial systems.