Lesson 13 of 22Go deeperbeginner4 min

History of AI

From the Turing Test to modern systems: 75+ years of progress

Why learn AI history?

AI did not appear overnight. The timeline below is a map of patterns: hype and winters, narrow wins that later broadened, architecture jumps that only became products after data and post-training caught up. Read milestones so you can pattern-match today's launch against past cycles and tell durable progress from marketing.

This is a teaching timeline

Parameter counts, release dates, and product names are simplified for a single-page story. Recent entries (2025-2026) describe product eras (reasoning tiers, computer use, on-device models) rather than one company's scoreboard. Throughlines are teaching patterns, not a full historiography. Click an event for detail; use era and category chips to focus.

Foundation

events

Breakthrough

events

Scale

events

Capability

events

Safety

events

Eras on this page

Foundations · 1950-1989(7 events)
Deep learning · 1990-2016(5 events)
Transformers · 2017-2021(5 events)
Mainstream · 2022-now(9 events)

26 milestones total · 1950-2026 · densest teaching era: Mainstream (9) · filters recompute from pure data in historyData

Throughlines: patterns that keep repeating

Pick a pattern. Example milestones come from the same timeline data the chart uses (matched by year). Read the lesson, then scan the full timeline for more context.

Hype, winter, spring

Pattern: Bold claims and funding waves, then disappointment when demos do not generalize, then a quieter technical fix that restarts the cycle.

What to do with it: Treat product launches as demos under constraints. Ask what broke last time (data, compute, evaluation, distribution) before declaring a new era permanent.

Winter/spring labels are teaching shorthand. Funding dips and research continuity were messier than a single timeline chip suggests.

Example milestones (4)

1956 Dartmouth Conference - AI is Born
John McCarthy, Marvin Minsky, Nathaniel Rochester, and Claude Shannon organize the Dartmouth Summer Research Project on Artificial Intelligence. This is where the term 'Artificial Intelligence' is coined and the field officially begins. They predicted significant progress within a summer.
1969 Perceptrons Book - First AI Winter
Minsky and Papert publish 'Perceptrons,' proving limitations of single-layer neural networks. While mathematically correct, the book's impact was overstated, leading to funding cuts and the first 'AI Winter': a period of reduced interest and investment in AI research.
2012 AlexNet wins ImageNet
A deep neural network called AlexNet crushes the ImageNet competition, cutting the error rate in half. This is the moment deep learning goes from academic curiosity to practical tool. The key? GPUs for training and massive labeled datasets.
2022 ChatGPT & Stable Diffusion
ChatGPT launches and reaches 100 million users in 2 months. Stable Diffusion brings open-source image generation. AI becomes mainstream overnight. The world changes. People who never cared about AI suddenly use it daily.

Era

Key Insight

The 2017 Transformer paper unlocked parallel attention over long contexts. Combined with scale, data, and later preference and reasoning post-training, that stack powers modern chat, coding, and multimodal products. Scale alone was not enough: architecture, data, post-training, and product distribution all had to line up. Use the throughlines above so the next launch is a pattern match, not a blank slate.

Check your understanding

1/5

What did Alan Turing propose in 1950?

Sources, scope & review statusVerified Jul 29, 2026

Claims this lesson makes

Stable conceptModern generative AI emerged from multiple research lineages rather than a single uninterrupted breakthrough.

Primary reading

Next content review: 2027-07-29. Source links establish the lesson's claims; interactive numbers remain labeled simulations unless explicitly identified as measured data.