François Chollet, the creator of Keras and a key figure in the AI world, is leaving Google after nearly a decade with the tech giant. The 34-year-old French developer shared the news on X, revealing that he’s starting a new company with a close collaborator.
Though he hasn’t shared many details, his departure marks a major shift in AI research and development, with Chollet’s ideas pushing the boundaries of what AI can achieve.
Chollet’s exit comes at a time when the AI landscape is undergoing rapid transformation. Keras, the deep learning framework he created, has become a cornerstone of modern AI development, used in everything from autonomous vehicles to personalized recommendations on platforms like YouTube and Spotify. While Chollet is moving on from Google, his vision for the future of AI—and his continued role in Keras—will have lasting effects on the industry.
Francois Chollet’s decision to leave Google comes as little surprise to those who follow his work closely. As the creator of Keras, Chollet has had an outsized influence on the development of deep learning technologies.
His contributions helped democratize AI by making complex machine learning models more accessible to developers worldwide. Google has acknowledged the profound impact Chollet has had, with the company expressing pride in his work and supporting his future endeavors.
Keras, which now boasts over two million users, powers critical systems in self-driving cars, content recommendations on YouTube and Netflix, and countless other applications. Despite his departure, Chollet will remain involved in Keras from the outside, ensuring its continued growth and supporting its open-source community. “I’m confident in the team’s ability to continue pushing the boundaries of what’s possible in deep learning,” he said in his farewell post.
Chollet’s ideas around AI have set him apart from many of his peers in the tech world.
While major companies like Google and OpenAI continue to push the limits of AI by feeding vast amounts of data into models, Chollet has long argued that this approach won’t lead to true artificial general intelligence (AGI).
He believes the key to developing AGI lies not in scaling up models, but in teaching them to reason in ways more similar to human thinking. To this end, he launched the $1.1 million ARC Prize in 2024, challenging researchers to solve reasoning problems that current AI systems struggle with.
So far, the results have validated his theory. While traditional AI approaches still lag behind humans in tasks that require reasoning, alternative methods—such as neuro-symbolic AI—are showing promise.
These approaches, which aim to replicate human-like reasoning, could eventually lead to AI systems that think more like people and less like complex calculators. Chollet sees AGI not as a super-charged chatbot, but as a tool capable of advancing human knowledge in profound ways. “Artificial general intelligence is going to be a kind of super-competent scientist,” he said in a recent interview.
Despite moving on from Google, Chollet’s departure is unlikely to slow down his impact. His ongoing work with Keras and the ARC Prize could play a pivotal role in shaping the future of AI.
Chollet’s focus on human-like reasoning is attracting a growing number of researchers and developers who are questioning the prevailing methods of scaling models. As AI continues to evolve, his influence is likely to drive a new wave of innovation, one that prioritizes intelligence over raw computational power.
While it’s still too early to say which direction the field will take, Chollet’s new venture is sure to be closely watched by those shaping the future of AI.
