Artificial intelligence did not appear overnight — it evolved through four distinct eras. The symbolic era began with Alan Turing’s 1950 paper “Computing Machinery and Intelligence” and tried to capture intelligence as hand-written rules: if X, then Y. It produced chess programs and expert systems, but rules written by humans could never cover the messiness of the real world.
The machine learning era flipped the approach: instead of writing rules, let the computer find patterns in examples. Algorithms like support vector machines (1992) learned decision boundaries from data, powering spam filters and recommendation systems — but humans still had to hand-design the features the models learned from.
In 2012, AlexNet’s win on the ImageNet challenge kicked off the deep learning era: neural networks with many layers that learn their own features directly from raw data, fueled by GPUs and massive datasets. Image recognition, speech recognition, and translation all jumped from “barely works” to “better than humans” within a few years.
The current generative era traces to a single 2017 paper, “Attention Is All You Need,” which introduced the transformer architecture. Trained on terabytes of text with one deceptively simple objective — predict the next word — transformers turned out to absorb grammar, facts, and reasoning patterns, producing the large language models behind ChatGPT, Claude, and Gemini. The interactive timeline on this page walks through all four eras, with the landmark paper from each one.
You just got the 30,000-foot view. Get my AI roadmap
howaiworks.io is free and open source (GitHub), built by Matt Feroz.