This paper introduces a novel framework for understanding cognitive attribution in AI systems, distinguishing between different ways AI can acquire and utilize representations. The framework, presented as a five-part attribution table, clarifies the scope of cognitive achievements, emphasizing that credit for established capacities should not implicitly grant broader, unproven repertoires. It is applied to analyze Othello-GPT's world-model interpretation and machine concepts, preserving recognition and classification while highlighting what remains undetermined about criterion discovery. AI
IMPACT Provides a theoretical framework for evaluating AI cognitive capabilities, potentially influencing future AI development and evaluation methodologies.
RANK_REASON Academic paper detailing a new framework for AI cognitive attribution. [lever_c_demoted from research: ic=1 ai=1.0]
- alphaXiv
- Andrew Ng
- arXiv
- CatalyzeX
- Connected Papers
- CORE Recommender
- DagsHub
- Gotit.pub
- Hugging Face
- Influence Flower
- Litmaps
- Othello GPT
- ScienceCast
- scite Smart Citations
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →