Journal of Artificial Intelligence and Consciousness
October 25, 2021
E.c. Garrido-Merchán, M. Molina, Francisco M. Mendoza-Soto
4 citations
Consciousness may offer evolutionary advantages that can be replicated in autonomous agents. This work presents a cognitive model for an autonomous agent based on global workspace theory, where a conscious-like bottleneck integrates and controls information from subsystems like attention, memory, and inner preferences. The agent navigates an environment of multiple independent magnitudes, adapting to find optimal positions according to its preferences. Large experiments show that agents with this architecture benefit from enhanced performance, supporting the idea that imitating a conscious cognitive structure can improve autonomous decision-making and adaptability.
Journal of Artificial Intelligence and Consciousness
September 1, 2024
Robin L. Zebrowski
This paper identifies common ground between Carlos Montemayor's account of mind and AI, as presented in his précis for *The Prospect of Humanitarian Artificial Intelligence*, and enactivist philosophy. The author notes that Montemayor emphasizes autonomy, embeddedness, perspective, and social interaction, yet does not engage with 4E (embodied, embedded, extended, enactive) cognitive science, particularly enactivism, despite his heavy reliance on biological and autonomous features to distinguish intelligence from consciousness.
Journal of Artificial Intelligence and Consciousness
October 31, 2023
Subhash C. Kak
Consciousness, often linked to free will, conflicts with the causal closure of physics, where every event has a physical cause. By analyzing nested physical systems, the paper argues that if a system had agency, observers could not exist within it. Since complex systems form nested hierarchies, this suggests consciousness cannot emerge from complexity alone. The existence of consciousness in cognitive agents implies it belongs to a non-physical dimension, making machine consciousness unattainable. These arguments are applied to reinterpret two quantum theory paradoxes relevant to quantum information theory.
Journal of Artificial Intelligence and Consciousness
September 1, 2020
Henry Shevlin
The science of consciousness has advanced but faces difficulty reaching consensus about artificial consciousness. Practical and ethical questions about whether artificial systems can suffer may soon arise, yet methods for assessing consciousness in humans and animals do not straightforwardly apply to artificial systems. The author proposes developing ecumenical heuristics for artificial consciousness that are intuitively plausible, theoretically neutral, and scientifically tractable. General intelligence—understood as robust, flexible, integrated cognition and behavior—is argued to satisfy these criteria and could provide a basis for making tentative assessments of which artificial systems are most likely to be conscious.