Turing proposes a practical alternative to the question "Can machines think?" by introducing the Imitation Game, later known as the Turing Test. This foundational paper frames the philosophical and technical boundaries of artificial intelligence, influencing decades of research in natural language processing, cognitive science, and machine learning. The paper also addresses common objections to machine intelligence, providing early arguments that remain relevant in modern AI discourse.
Raphael, B. — *Artificial Intelligence*, North-Holland
This paper formalizes the structure and inference mechanisms of semantic networks, laying groundwork for knowledge representation systems. It introduces inheritance, polymorphism, and cross-linked node structures that would later influence object-oriented programming and modern knowledge graphs. The author demonstrates how semantic networks can model hierarchical relationships and support non-monotonic reasoning.
This landmark paper introduces PageRank, an algorithm that quantifies the importance of web pages through link analysis. By modeling the web as a directed graph and applying iterative probability distributions, Page and Brin solved the problem of ranking search results by relevance rather than keyword frequency. The paper details implementation at scale, caching strategies, and trust metrics that power modern search infrastructure.
Krizhevsky, A., Sutskever, I., & Hinton, G. — *NeurIPS*, MIT Press
The AlexNet paper demonstrates a breakthrough in visual recognition by training a deep convolutional neural network on ImageNet. Utilizing GPU acceleration and ReLU activations, the model achieved unprecedented accuracy, catalyzing the modern deep learning revolution. The paper details architecture design, data augmentation techniques, and dropout regularization, establishing standards still used in computer vision today.
This paper introduces the Transformer architecture, replacing recurrent and convolutional structures with self-attention mechanisms. By parallelizing sequence processing and capturing long-range dependencies, Transformers dramatically improved translation, text generation, and multimodal tasks. The model's scalability and efficiency established the foundation for modern large language models and generative AI systems.