# Definitions and a Short History — Artificial Intelligence

Source: https://www.skillbyai.com/en/artificial-intelligence/i-what

> From logic machines to learning systems.

## Seventy years of ideas

**Artificial intelligence (AI)** is the field that builds systems able to perform tasks we associate with intelligence: reasoning, planning, learning, perceiving, understanding language. The term dates from the 1956 Dartmouth workshop. Early AI focused on **symbolic** methods: search, logic and hand-written rules, leading to expert systems in the 1980s. Periods of over-promising were followed by funding cuts ("AI winters"). From the 1990s, **probabilistic** methods and **machine learning** took over; since the 2010s, **deep learning** on large data and compute has driven breakthroughs in vision, speech, games and, recently, large language models.

## Machines that act sensibly

AI studies how to build systems that perceive, reason, learn and act to achieve goals.

![Three ideas: definitions and history, agents, kinds of AI.](assets/figures/artificial-intelligence/section-1-map.svg) — Figure 1.1 — History, agents and kinds of AI.

## Milestones

A compressed timeline.

```text
1950  Turing asks "Can machines think?" and proposes the imitation game
1956  Dartmouth workshop names the field "artificial intelligence"
1960s-80s  search, logic, expert systems; AI winters follow over-promising
1997  Deep Blue defeats the world chess champion (search + evaluation)
2012  deep neural networks win ImageNet by a large margin
2016  AlphaGo beats a top Go professional (deep learning + search + RL)
2020s large language models and generative AI reach mainstream use
```

## Old ideas still matter

Search, logic and probability are inside modern systems too: game engines, planners, route finders and LLM agents all use them.

**Quiz:** Which approach dominated early AI research?

- [ ] Deep convolutional networks
- [ ] Large language models
- [x] Symbolic methods: search, logic and hand-written rules
- [ ] GPU clusters

*Answer:* Symbolic methods: search, logic and hand-written rules. Learning-based methods rose later.