Ian Goodfellow

Co-founder & CEO — Elorian AI Research Scientist — Google DeepMind Director of Machine Learning — Apple Research Scientist — Google Brain

Listen — profile

0:00 / 1:54

Biographies

Genius Makers

The Mavericks Who Brought AI to Google, Facebook, and the World

Cade Metz 2021

NYT journalist Cade Metz's account of the AI pioneers — Hinton, LeCun, Bengio, Goodfellow — who brought deep learning from academia to tech giants.

Genius Makers: The Mavericks Who Brought A.I. to Google, Facebook, and the World

Cade Metz 2021

Journalist Cade Metz's history of AI pioneers — Goodfellow's GAN breakthrough at Montreal bar to Google.

Profile

Ian Goodfellow invented Generative Adversarial Networks (GANs) in 2014, reportedly while arguing with colleagues at a bar in Montreal. The idea: pit two neural networks against each other — one generating fakes, one trying to spot them — and let them co-evolve until the fakes are indistinguishable from real data. He went home, coded it up that night, and it worked on the first try. That paper kicked off the entire field of generative AI, and although diffusion models have since taken over image synthesis, every “AI-generated face” meme of the last decade traces back to that one night.

He did his PhD at Université de Montréal under Yoshua Bengio and co-authored the 2016 textbook Deep Learning with Bengio and Aaron Courville — the book most practitioners just call “the Goodfellow book.” Before GANs he was one of the earliest hires at OpenAI, then spent years at Google Brain leading adversarial ML work, before jumping to Apple as Director of Machine Learning in 2019.

In May 2022 he quit Apple over its return-to-office mandate, in what became one of the most public pushbacks against post-pandemic RTO policies in big tech. He landed at Google DeepMind as a research scientist, where he remains.

Why he matters to anyone building with AI today: beyond GANs, Goodfellow essentially created modern adversarial machine learning as a field. His work on adversarial examples — tiny imperceptible perturbations that fool neural networks — is the reason every ML security curriculum starts where it does. If you care about model robustness, why systems can be fooled, or the theoretical roots of today’s image generators, you are standing on his papers.

Books

Deep Learning

2016

The definitive textbook on deep learning, co-authored with Yoshua Bengio and Aaron Courville — free online and still a standard reference years after publication.

Key Articles & Papers

Generative Adversarial Networks 2014 — The original GAN paper. Launched generative AI as a field and remains one of the most-cited ML papers ever written. Explaining and Harnessing Adversarial Examples 2014 — Introduced the fast gradient sign method and explained why neural networks are fragile to tiny worst-case perturbations. NIPS 2016 Tutorial: Generative Adversarial Networks 2016 — Goodfellow's own written companion to his famous NIPS 2016 GAN tutorial — still one of the clearest introductions to how GANs actually work.

YouTube

2022

2017

2015

Spotify Podcasts

Ep 15: Ian Goodfellow designed GANs in 2014 to let machines create realistic images from scratch; five years later the same system was generating non-consensual pornography and fabricated political videos at scale.

Unintended Consequences

2026