1956 · A shared idea
The museum
In 1956, the Dartmouth meeting helped establish AI as a research field. Earlier work already existed, and much remained to be discovered.
DartmouthThese sources explain the journey’s milestones. Missions are fictional scenes and educational simplifications, not exact reproductions of historical systems. Editorial review: September 2026.
1956 · A shared idea
In 1956, the Dartmouth meeting helped establish AI as a research field. Earlier work already existed, and much remained to be discovered.
Dartmouth1966 · Early conversations
Joseph Weizenbaum’s ELIZA used rules and words from a message to build responses. It worked differently from today’s generative assistants.
Computer History Museum1997 · A specific task
In 1997, Deep Blue defeated Garry Kasparov in a chess match under tournament controls. The victory demonstrated a specialised chess capability.
IBM2012 · Learning from examples
The work known as AlexNet achieved a major image-classification result in 2012. Learning from examples and image recognition had earlier precedents. Here we use a simple simulation.
NeurIPS2016 · Improving through practice
AlphaGo beat Lee Sedol 4–1 in 2016. It combined learning and search to play Go. The delivery exercise simplifies the idea of testing a strategy and observing results.
Google DeepMind2016 · What gets recommended
A 2016 Google paper described YouTube recommendation systems. Recommendations already existed. Our simulated market explores how activity can influence what gets shown.
Google Research2022 · Asking through conversation
ChatGPT was introduced on 30 November 2022. Its conversation format allowed requests and follow-up changes. Artificial conversation and text generation had earlier precedents.
OpenAI2023 · Images and conversation
In September 2023, a gradual rollout of image and voice in ChatGPT was announced. It is an example of how ways of interacting with these assistants expanded.
OpenAI2024 · Seeing and hearing in context
In 2024, developments such as GPT-4o combined text, images and audio. Suspected voice impersonation attempts were also documented. Appearances or a familiar voice need context to verify their origin.
INCIBEOpenAI2025 · Tools that act
Operator was introduced in January 2025 as a research preview able to act in a browser. Earlier work included Anthropic’s 2024 computer-use announcement. These capabilities need limits and review.
OpenAIAnthropicToday · Start here
AI appears in everyday tasks. This journey uses prepared examples to help you practise and check results.
Fundación Telefónica2026 · What you can do today
September 2026 is the destination of this journey. The year is still unfolding. Here we apply what we learned without presenting future predictions as facts.
CHI EA 2025