Skip to content
Deepak Gupta markBooks

Books

The Worlds I See

Editorial pick

Curiosity, Exploration, and Discovery at the Dawn of AI

By Fei-Fei Li · Flatiron Books · 2023

The first major AI memoir from a working researcher rather than a CEO or venture capitalist.

Memoir 200–350 pages(336p) Beginner Published 2023
Amazon Publisher Goodreads Author site

Editorial take

Most AI books written for general audiences are either breathless ("how AI will change everything") or panicked ("how AI will end everything"). Li's memoir is neither. She writes from inside the field as the researcher who created ImageNet — the dataset that, more than any single paper, enabled the deep learning revolution. The book is half immigrant coming-of-age story and half history of computer vision; both halves work. Required reading for anyone trying to understand modern AI as a scientific tradition rather than a 2023-era product launch. Pair with any of the popular AI policy books for an unusually grounded picture.

Last hand-checked 2026-05-18.

Read if you …

  • build with AI/ML and want the actual historical context of why deep learning won
  • are tired of AI books written by people who've never trained a model
  • want the ImageNet origin story from the person who built it

Skip if you …

  • you want a technical textbook — the technical content is general-audience accessible
  • you wanted a policy/governance book — this is memoir-and-history

If you only read one chapter

ImageNet's Decade

Li's account of the eight-year ImageNet effort — when nobody else thought the data scale mattered — is the single most important business case for patient research investment in print.

Key ideas

  • Data scale, not algorithmic novelty, broke through computer vision's plateau.
  • Most fundamental breakthroughs require sustained, unfashionable investment.
  • Human-centered AI is a research agenda, not a marketing line.
  • The most underrated input to scientific progress is a good benchmark dataset.

About the book

Fei-Fei Li — Stanford professor, former Chief Scientist of AI/ML at Google Cloud, current co-director of Stanford HAI — wrote The Worlds I See as a dual narrative: her own immigration from China to the United States as a teenager, and the parallel history of computer vision from the 1990s through the deep-learning era she helped enable.

The ImageNet chapters are the book's intellectual core. Li was the principal architect of a dataset whose scale (14 million labeled images) at the time looked absurd, and which retroactively turned out to be the single most enabling resource for the deep-learning breakthroughs of the 2010s.

Pairs with

If The Worlds I See works for you, these likely will too.