History of Artificial Intelligence - From Symbolic Logic to Deep Learning

Artificial intelligence has gone through repeated cycles of optimism, disappointment, and resurgence since its formal founding, each driven by a shift in the dominant technical approach.

The Founding Era (1950s-1960s)

The field’s name was coined at the 1956 Dartmouth Workshop, where researchers proposed that “every aspect of learning… can in principle be so precisely described that a machine can be made to simulate it.” Early systems like the Logic Theorist and General Problem Solver used symbolic search and hand-crafted rules to solve puzzles and prove theorems, fueling confident predictions that human-level intelligence was only a couple of decades away.

The First AI Winter (1970s)

Symbolic approaches struggled with real-world ambiguity and combinatorial explosion—the number of possibilities to search grew too fast as problems became realistic. Funding bodies, notably reacting to critical reports like the UK’s Lighthill Report, cut support sharply, leading to the first “AI winter”: a period of reduced funding and interest.

Expert Systems and the Second Winter (1980s-early 1990s)

AI resurged with expert systems—programs encoding domain-specific rules from human experts (e.g., MYCIN for medical diagnosis). These found real commercial use but were brittle, expensive to maintain, and could not generalize beyond their hand-coded rules. Combined with the collapse of the specialized “Lisp machine” hardware market, this led to a second winter by the late 1980s.

The Statistical and Machine Learning Turn (1990s-2000s)

Interest shifted from hand-coded rules to statistical methods that learned from data: support vector machines, decision trees, and probabilistic graphical models. Milestones like IBM’s Deep Blue defeating chess champion Garry Kasparov in 1997 demonstrated the power of search combined with domain-specific evaluation, without any claim to general intelligence.

The Deep Learning Resurgence (2012-present)

Three ingredients converged: large labeled datasets (ImageNet), GPU-accelerated computation, and improved neural network training techniques. AlexNet’s decisive win in the 2012 ImageNet competition marked the beginning of deep learning’s dominance in computer vision, quickly followed by breakthroughs in speech recognition, machine translation, and game playing (AlphaGo, 2016).

The Transformer and Large Model Era (2017-present)

The 2017 “Attention Is All You Need” paper introduced the Transformer architecture, which became the foundation for large language models. Scaling these models in parameters and training data, combined with techniques like reinforcement learning from human feedback, produced the general-purpose conversational AI systems widely used today. Whether this trajectory continues smoothly or encounters another slowdown remains an open and actively debated question.