François Chollet, the renowned AI researcher and creator of the Keras deep learning library, has officially left Google after over nine years with the company. His departure marks a significant moment in the AI community, given his influence in the field of machine learning and artificial intelligence. During his time at Google, Chollet contributed to advancements in computer vision, machine learning, and benchmarks for artificial general intelligence (AGI), such as the Abstraction and Reasoning Corpus (ARC).
In recent years, Chollet has been vocal about his perspectives on the limitations of current AI, especially large language models (LLMs), which he views as insufficient for achieving AGI. He argues that intelligence involves more than pattern recognition or memorization, emphasizing adaptability and reasoning as core components. To push the boundaries of AI research, he launched the ARC Prize, a $1 million competition designed to encourage innovative approaches toward AGI.
François Chollet Paper
François Chollet’s ARC Prize initiative, launched in June 2024, aims to spur advancements in AI reasoning by focusing on solving tasks from the Abstraction and Reasoning Corpus (ARC). This competition builds on the ARC benchmark he introduced in 2019, which tests an AI’s ability to solve novel reasoning problems.
Early AI systems could only solve 20% of ARC tasks, and by mid-2024, improvements raised this to 34%, still trailing human performance of 84%. To encourage breakthroughs, Chollet partnered with Mike Knoop, co-founder of Zapier, offering a $1 million prize pool. Participants must develop systems capable of tackling 400 ARC tasks in public and private evaluations.
François Chollet’s departure from Google appears to stem from both professional and philosophical differences concerning the direction of AI development. He has described current AI trends as overly focused on scale and data memorization rather than innovation in reasoning and adaptability.
Additionally, Chollet has expressed concerns about corporate secrecy in AI research. His launch of the ARC Prize reflects a commitment to open-source solutions and a desire to foster community-driven progress in AI. His decision to leave could also indicate a shift toward independent pursuits. That is advancing the ARC Prize competition and continuing his research on AGI in less restrictive environments.
