From Dartmouth in 1956, through expert systems and statistical ML, to deep learning and large language models.
From the chapter: Chapter 1: What Is AI?
Glossary: dartmouth workshop, expert systems
Transcript
In 1956, the Dartmouth Workshop is convened by McCarthy, Minsky, Shannon, and Rochester. The phrase artificial intelligence is coined there. The first wave begins on a tide of optimism.
1960s and 1970s. Symbolic systems. Logic theorem provers, planners, game-playing programs. Bold predictions of full intelligence within a decade. By 1974, results disappoint and funding collapses. The first AI winter.
1980s. Expert systems. Hand-coded rules in narrow domains, medical diagnosis, configuration. A second commercial wave. By the late eighties, brittleness exposed. Second winter.
1990s and early 2000s. Statistical machine learning. Support vector machines, random forests, graphical models. Steady progress, modest claims. Speech and translation begin to work.
AlexNet wins ImageNet by ten percentage points. Deep learning takes over computer vision. Then speech, then language. The third wave gathers momentum.
The transformer architecture is introduced. By 2020, GPT-3 shows few-shot learning. By 2022, ChatGPT reaches a hundred million users in two months.
2024 onwards. Reasoning models. Multimodal systems. Scientific tools and agents. New waves now arrive faster than the winters between them.
Every wave inherits from the one before it. The seventh begins now, and that is the subject of this book.