François Chollet
In plain wordsA French software engineer who created Keras, a popular tool for building neural networks, and the ARC benchmark, which tests whether AI can solve kinds of puzzles it has never seen.
French software engineer and AI researcher who created KerasKeras, prop. n.: A free, beginner-friendly Python library for building neural networks from ready-made layers in a few lines of code, while a larger toolkit underneath does the heavy computing. (2015), a high-level library that made building neural networksneural network, n.: A kind of computer program that learns a task from examples by tuning thousands to billions of connections between simple number-crunching units, an arrangement loosely inspired by nerve cells in the brain. accessible and became TensorFlowTensorFlow, prop. n.: Google's free software toolkit for building, training and running machine-learning models; it led the field after 2015 and still runs many models inside phones, browsers and company systems.'s official interface, and the Xception architecture (2016), built on depthwise separable convolutionsdepthwise separable convolution, n.: A cheaper way to do the main operation of image networks: first scan each channel of the input separately, then mix the channels together with a tiny one-by-one step.. His paper 'On the Measure of Intelligence' (2019) argued that intelligence should be measured as skill-acquisition efficiency — how well a system generalizes to novel tasks from little data, given its prior knowledge — rather than as skill at known tasks, and introduced the Abstraction and Reasoning Corpus (ARC-AGIARC-AGI, prop. n.: Puzzles made of colored grids — and, since 2026, small video-game-like worlds — that people solve easily but AI long found hard, designed to test whether a machine can learn a new skill from just a few examples.), grid puzzles easy for people but long resistant to AI. A critic of the view that scaling language modelslarge language model, n.: A computer program trained on enormous amounts of text to predict the words that come next. At vast scale, that simple skill lets it chat, answer questions, translate, summarize and write code. suffices for general intelligence, he launched the ARC Prize with Mike Knoop in 2024. He left Google, where he had worked since 2015, in late 2024, and in 2025 co-founded with Knoop the research lab Ndea, which combines deep learning with program synthesisprogram synthesis, n.: Getting a computer to write a program automatically from a description of what the program should do — a logical specification, a few input–output examples or a request in plain language..
Related
- ARC-AGIARC-AGI, prop. n.: Puzzles made of colored grids — and, since 2026, small video-game-like worlds — that people solve easily but AI long found hard, designed to test whether a machine can learn a new skill from just a few examples.
- KerasKeras, prop. n.: A free, beginner-friendly Python library for building neural networks from ready-made layers in a few lines of code, while a larger toolkit underneath does the heavy computing.
- intelligenceintelligence, n.: The ability to learn, understand, reason and solve problems — especially new ones — in order to reach goals. Experts disagree about how to define it precisely, for people and even more for machines.
- program synthesisprogram synthesis, n.: Getting a computer to write a program automatically from a description of what the program should do — a logical specification, a few input–output examples or a request in plain language.
- benchmarkbenchmark, n.: A fixed, shared test — a set of questions or tasks with an agreed way of scoring — used to compare AI systems with one another and to track progress over time.
- depthwise separable convolutiondepthwise separable convolution, n.: A cheaper way to do the main operation of image networks: first scan each channel of the input separately, then mix the channels together with a tiny one-by-one step.
Sources
- Chollet, F. (2019). On the Measure of Intelligence. arXiv preprint.
- Chollet, F. (2017). Xception: Deep Learning with Depthwise Separable Convolutions. CVPR 2017.