Abstract

Artificial General Intelligence (AGI) refers to computational systems capable of performing intellectual tasks at human‑like levels across diverse domains, including reasoning, decision‑making, learning, and problem‑solving. AGI architectures aim to integrate perception, cognition, language, and action within a unified system. Achieving AGI requires innovations beyond narrow AI paradigms that dominate current technology. Contemporary research examines cognitive architectures, hybrid symbolic‑connectionist systems, developmental and embodied approaches, and modular frameworks that mimic human cognitive structures. Significant design challenges include knowledge representation, scalable learning, generalization across contexts, explainability, safety, and ethics. Opportunities abound through hybrid architectures, agentic systems, neuromorphic hardware, and integrated human‑machine collaboration. Hybrid cognitive design patterns combining deep learning with symbolic reasoning, multi‑agent coordination, and universal knowledge models present promising pathways toward AGI. However, limitations in computational resources, task transferability, and alignment with human values represent considerable impediments. This paper synthesizes current AGI architectural approaches, identifies core design challenges, evaluates opportunities for advancement, and highlights future research trajectories. By clarifying the landscape of AGI architecture research and grounding discussions in both classical foundations and modern developments, this work aims to provide a comprehensive platform for advancing AGI design and deployment. ScienceDirect+1