LLMilestonesA milestone timeline of AI and large language models

Prologue · Before the Transformer

PaperKrizhevsky, Sutskever & Hinton

AlexNet

A deep convolutional network trained on two GPUs that won the ImageNet image-recognition challenge by a huge margin.

15.3%top-5 error, vs 26.2% for second place

Why it matters

Its error rate was about 11 points lower than the runner-up, and deep learning went mainstream. Co-author Ilya Sutskever later co-founded OpenAI.

Built on

BackpropLeNet

Led to

DQNGANTransformer

Source: papers.nips.cc ↗

See it on the 3D timeline →