| The Epistemological Architectures of Artificial Intelligence |
| کد مقاله : 1078-CYSP2025 (R1) |
| نویسندگان |
|
رامین رامبد *، هادی صمدی، شهلا اسلامی گروه فلسفه واحد علوم و تحقیقات دانشگاه آزاد اسلامی |
| چکیده مقاله |
| This article provides a comprehensive analysis of how different artificial intelligence paradigms approach and represent truth across three key epistemological dimensions: knowing-that (declarative), knowing-what (conceptual), and knowing-how (procedural). The historical schism between Symbolic and Connectionist AI is an explicit computational manifestation of an underlying philosophical divide. Symbolic AI, with its reliance on explicit rules and structured data, excels at representing truth in a verifiable, but often brittle, manner. Conversely, Connectionist AI, through implicit representations in neural networks, achieves impressive generalization but sacrifices interpretability. The emergent Neuro-Symbolic paradigm represents a conceptual synthesis, attempting to merge the strengths of both approaches to create systems that are simultaneously adaptive, explainable, and robust. The nature of "truth" in AI is not a singular concept but is fundamentally dependent on the system's architecture. The AI paradigms are not subjective engineering choices but are the direct result of assuming a fundamentally epistemological stance. We conclude that a complete, general-purpose intelligence may require a hybrid framework that can seamlessly integrate the explicit logic of symbolic systems with the emergent, probabilistic knowledge of neural networks, mirroring the complexities of human cognition. |
| کلیدواژه ها |
| Epistemology, Symbolic AI, Connectionist AI, Neuro-Symbolic AI, Knowledge Representation, AGI. |
| وضعیت: پذیرفته شده برای ارائه شفاهی |