François Chollet

francois.chollet@gmail.com

Software engineer and AI researcher. Co-founder Ndea, ARC Prize.

Primary interests:

Understanding the nature of abstraction and developing algorithms capable of autonomous abstraction (i.e. general intelligence)
Democratizing the development and deployment of AI technology, by making it easier to use and explaining it clearly
Leveraging technology, in particular AI, to help people gain greater agency over their circumstances and reach their full potential (e.g. EdTech, steerable recommendation engines, personal productivity tech, etc.)
Understanding and simulating the early stages of human cognitive development (e.g. developmental psychology, cognitive developmental robotics)

Books

Deep Learning With Python, Second Edition

F. Chollet

2021

In this revised and expanded new edition, Keras creator François Chollet offers insights for both novice and experienced machine learning practitioners. As you move through this book, you’ll build your understanding through intuitive explanations, crisp color illustrations, and clear examples. You’ll quickly pick up the skills you need to start developing deep-learning applications.

Publisher link

Deep Learning With Python, First Edition

F. Chollet

2017

Deep Learning with Python introduces the field of deep learning using the Python language and the powerful Keras library. This book builds your understanding through intuitive explanations and practical examples. You'll explore challenging concepts and practice with applications in computer vision, natural-language processing, and generative models. By the time you finish, you'll have the knowledge and hands-on skills to apply deep learning in your own projects.

Publisher link

Translations

German

Chinese (simplified)

Japanese

Korean

Spanish

French

Turkish

Polish

Chinese (traditional)

Russian

Italian

Czech

Deep Learning With R

F. Chollet, J. J. Allaire

2018

Deep Learning With R is the R language version of Deep Learning With Python.

Publisher link

Software

Keras - Deep learning library

2015

Creator, project lead, 2015-present.

TensorFlow - Machine learning platform

2015

Contributor, 2015-2021.

Keras Tuner - Hyperparameter tuning for Keras

2019

Project lead, contributor, 2019-present.

AutoKeras - Automated machine learning for Keras

2018

Contributor, 2019-present.

Wysp - Social network and learning engine

2011

Creator, maintainer, 2011-2014.

Wysp is a social network for artists, featuring tools to learn to draw. It peaked at 25,000 MAU around 2014-2015.

QuickAnswers.io (retired)

2014

Creator, maintainer, 2014-2015.

QuickAnswers was an online question-answering engine and dialogue system. It was similar in UX to ChatGPT. It used a technique that would later become known as RAG (Retrieval-Augmented Generation): Given a query, it would normalize it into a standard form, run a Google search, extract text relevant snippet from the top results, and pass these snippets to a LSTM language model that would generate a context-appropriate user-facing response.

Papers

ARC Prize 2024: Technical Report

2024

F. Chollet, M. Knoop, G. Kamradt, B. Landers

AutoKeras: An AutoML Library for Deep Learning

2020

H. Jin, F. Chollet, Q. Song, X. Hu

JMLR

Forecasting SARS-CoV-2 transmission and clinical risk at small spatial scales by the application of machine learning architectures to syndromic surveillance data

2020

T. Ward, A. Johnsen, S. Ng, F. Chollet

Nature Machine Intelligence

On the Measure of Intelligence

2019

F. Chollet

Tensor2Tensor for Neural Machine Translation

2018

A. Vaswani, S. Bengio, E. Brevdo, F. Chollet, A. N. Gomez, S. Gouws, L. Jones, L. Kaiser, N. Kalchbrenner, N. Parmar, R. Sepassi, N. Shazeer, J. Uszkoreit

Depthwise Separable Convolutions for Neural Machine Translation

2017

L. Kaiser, A. N. Gomez, F. Chollet

ICLR

Holstep: A Machine Learning Dataset for Higher-Order Logic Theorem Proving

2017

C. Kaliszyk, F. Chollet, C. Szegedy

ICLR

Xception: Deep learning with Depthwise Separable Convolutions

2016

F. Chollet

CVPR

Information-Theoretical Label Embeddings for Large-Scale Image Classification

2016

F. Chollet

DeepMath: Deep Sequence Models for Premise Selection

2016

A. A. Alemi, F. Chollet, N. Een, G. Irving, C. Szegedy, J. Urban

NeurIPS

An Information-Theoretical Approach to Unsupervized Learning Over Data Streams

2012

F. Chollet, Y. Kuniyoshi

Tracing Commodities in Indoor Environments for Service Robotics

2011

O. M. Mozos, F. Chollet, K. Murakami, K. Morooka, T. Tsuji, R. Kurazume, T. Hasegawa

IFAC Symposium on Robot Control

Essays

How I think about LLM prompt engineering

October 9, 2023

Prompting as searching through a space of vector programs.

The Memories Around Us

December 30, 2018

Every place has a memory. Some places are a memory, a record, layer after layer of history.

Notes to Myself on Software Engineering

September 8, 2018

Design for ethics. Bake your values into your creations.

What Worries Me About AI

March 28, 2018

Don't use AI as a tool to manipulate your users; instead, give AI to your users as a tool to gain greater agency over their circumstances.

The Implausibility of Intelligence Explosion

November 27, 2017

The notion of intelligence explosion comes from a profound misunderstanding of both the nature of intelligence and the behavior of recursively self-augmenting systems.

User Experience Design for APIs

November 21, 2017

Like most things, API design is not complicated, it just involves following a few basic rules. They all derive from a founding principle: you should care about your users.

The Future of Deep Learning

July 18, 2017

We will move away from having on one hand "hard-coded algorithmic intelligence" (handcrafted software) and on the other hand "learned geometric intelligence" (deep learning). We will have instead a blend of formal algorithmic modules that provide reasoning and abstraction capabilities, and geometric modules that provide informal intuition and pattern recognition capabilities. The whole system would be learned with little or no human involvement.

The Limitations of Deep Learning

July 17, 2017

Despite our progress on machine perception, we are still very far from human-level AI: our models can only perform local generalization, adapting to new situations that must stay very close from past data, while human cognition is capable of extreme generalization, quickly adapting to radically novel situations, or planning very for long-term future situations.

On the Importance of Democratizing Artificial Intelligence

July 6, 2016

It's not a given that every technological revolution should turn out as a net positive for humanity, empowering individuals and bringing us higher potential for learning and creating, for self-direction and self-actualization. It's our responsibility to make sure that this new revolution turns out all right.

The Singularity Is Not Coming

August 10, 2012

Despite exponential resource investment, the pace of scientific progress has been linear.

The Good Librarian

April 30, 2011

The web has become a huge distraction. But it doesn't have to be that way.

The Piano-Playing Cat Paradigm

December 5, 2010

Our use of the Internet is conditioned by its infrastructure. By the underlying logic of current social networks and browsing tools.