Rishi Siddharth

A Timeline of AI

1950 – 2026 · people, papers, hardware, events

Artificial intelligence took seventy-six years to go from a thought experiment to autonomous agents. The ideas came early. What arrived late was the compute and data to make them work. This page follows that path through the papers and machines that mattered most.

~109×more compute to train frontier models than AlexNet (2012)
~7 modoubling time of the tasks AI agents can finish alone
900Mweekly ChatGPT users, February 2026

Photos and diagrams: Wikimedia Commons; each caption links to the file page with its author and license. Charts: Rishi Siddharth, from data by Epoch AI, METR, OpenAI, DeepSeek, Nvidia and the ImageNet challenge.

← Back home