Tag: #ai-theory5 results
MIT's Tomaso Poggio likens current AI to pre-Maxwell electricity, driven by engineering without deep theory. He advocates sparse compositionality—combining simple low-dim functions—for generalization in intelligent systems. Discussion covers his learning theories from kernel machines to deep nets.
This paper introduces Causal Reinforcement Learning (CRL), a framework that bridges causal inference and reinforcement learning. By modeling environments as structural causal models, it enables agents to reason about counterfactuals and optimize policies more effectively.
This position paper argues that AI research must shift from analyzing static models to understanding the time-evolving training dynamics. It proposes a scientific framework to predict model behaviors and design more reliable training procedures.
This article explores the historical evolution of probability theory and its role in shaping modern scientific rationality. It highlights the limitations of Bayesian inference in handling conceptual vagueness and discusses how fuzzy logic and deep learning complement traditional probabilistic models.
AI traditionally compared to human intelligence may be misguided. 'Jagged intelligence' highlights AI's uneven strengths and weaknesses. This framework better predicts jobs AI could replace.