Machine learning · Neural architectures · Natural language processing
The earliest counting boards appear in Mesopotamia, providing a tangible tool for arithmetic that would influence computational aids for millennia across civilizations. · Wikipedia: https://en.wikipedia.org/wiki/Abacus
The Indian scholar Panini produces the Ashtadhyayi, a rigorous formal grammar of Sanskrit with production rules that anticipate modern formal-language theory and computational linguistics. Source — Wikipedia:
Counting rods are used in China for positional decimal arithmetic and solving linear equations, representing one of the earliest systematic mechanical aids for mathematical computation. Source — Wikipedia:
Persian scholar al-Khwarizmi writes Al-Jabr, formalizing algebraic methods of solving equations; his name gives us the word 'algorithm,' foundational to all computer science. · Wikipedia: https://en.wikipedia.org/wiki/Muhammad_ibn_Musa_al-Khwarizmi
Arab engineer al-Jazari designs a musical automaton with adjustable camshafts, an early example of programmable mechanical control described in his Book of Knowledge of Ingenious Mechanical Devices. · Wikipedia: https://en.wikipedia.org/wiki/Ismail_al-Jazari
Alan Turing publishes 'On Computable Numbers,' defining the universal Turing machine and establishing the mathematical foundations for all general-purpose computation. · Wikipedia: https://en.wikipedia.org/wiki/Turing_machine
Warren McCulloch and Walter Pitts propose the first mathematical model of a biological neuron, showing that networks of such units can in principle compute any logical function. · Wikipedia: https://en.wikipedia.org/wiki/McCulloch%E2%80%93Pitts_neuron
Alan Turing publishes 'Computing Machinery and Intelligence,' introducing the Turing Test as a criterion for machine intelligence and framing the question 'Can machines think?' · Wikipedia: https://en.wikipedia.org/wiki/Turing_test
John McCarthy, Marvin Minsky, Claude Shannon, and Nathaniel Rochester organize the Dartmouth Summer Research Project, coining the term 'artificial intelligence' and establishing it as a field. Source — Wikipedia:
Frank Rosenblatt introduces the perceptron, the first trainable neural network model, enabling machines to learn binary classification from data. Source — Wikipedia:
Joseph Weizenbaum at MIT builds ELIZA, a natural language processing program that simulates a Rogerian psychotherapist and demonstrates surprisingly human-like conversational ability. · Wikipedia: https://en.wikipedia.org/wiki/ELIZA
Marvin Minsky and Seymour Papert publish 'Perceptrons,' proving single-layer networks cannot solve XOR and contributing to reduced neural network funding for years. Source — Wikipedia: )
Kunihiko Fukushima introduces the Neocognitron, a hierarchical multilayer neural network inspired by visual cortex that is a direct precursor to convolutional neural networks. · Wikipedia: https://en.wikipedia.org/wiki/Neocognitron
David Rumelhart, Geoffrey Hinton, and Ronald Williams publish the backpropagation learning algorithm, enabling effective training of multi-layer neural networks and reviving the field. · Wikipedia: https://en.wikipedia.org/wiki/Backpropagation
IBM's Deep Blue defeats Garry Kasparov, marking the first time a machine beats a reigning world chess champion under tournament conditions and a milestone in game-playing AI. Source — Wikipedia: )
Sepp Hochreiter and Jürgen Schmidhuber introduce Long Short-Term Memory networks, solving the vanishing gradient problem and enabling recurrent networks to learn long-range dependencies. · Wikipedia: https://en.wikipedia.org/wiki/Long_short-term_memory
IBM's Watson defeats human champions Ken Jennings and Brad Rutter on Jeopardy!, demonstrating advanced natural language understanding and question-answering at scale. Source — Wikipedia:
Alex Krizhevsky, Ilya Sutskever, and Geoffrey Hinton's AlexNet wins the ImageNet competition by a large margin, igniting the deep learning revolution through GPU-trained convolutional networks. · Wikipedia: https://en.wikipedia.org/wiki/AlexNet
Ian Goodfellow and colleagues invent Generative Adversarial Networks, pitting generator against discriminator to produce realistic synthetic data and advancing generative modeling. · Wikipedia: https://en.wikipedia.org/wiki/Generative_adversarial_network
DeepMind's AlphaGo defeats world champion Lee Sedol in Go, a game long thought resistant to AI, using deep reinforcement learning and Monte Carlo tree search. · Wikipedia: https://en.wikipedia.org/wiki/AlphaGo
Ashish Vaswani and colleagues at Google publish 'Attention Is All You Need,' introducing the Transformer architecture that enables parallel training and revolutionizes natural language processing. · Wikipedia: https://en.wikipedia.org/wiki/Transformer_(deep_learning_architecture)
Jacob Devlin and colleagues at Google release BERT, a bidirectional Transformer model that sets new state-of-the-art results across natural language understanding benchmarks. · Wikipedia: https://en.wikipedia.org/wiki/BERT_(language_model)
OpenAI releases GPT-3, a 175-billion-parameter language model that demonstrates remarkable few-shot learning across diverse natural language tasks, showcasing emergent capabilities of scale. · Wikipedia: https://en.wikipedia.org/wiki/GPT-3
OpenAI launches ChatGPT, a conversational interface to GPT-3.5 that reaches 100 million users in two months and makes large language models a mainstream technology. Source — Wikipedia:
OpenAI releases GPT-4, a multimodal model that accepts both text and images while achieving human-level performance on many professional and academic benchmarks. Source — Wikipedia: