Francois Chollet: I would not "declare AGI" until we have AI that is capable of invention, conceptual breakthroughs, novel insights, new real-world technology, etc.

Francois Chollet argues against declaring AGI until AI can achieve invention, conceptual breakthroughs, novel insights, and new real-world technology. He notes that the original narrative for AGI focused on solving major problems like cancer or fusion energy. Chollet suggests that those eager to label systems like GPT-6 Astra as AGI resemble salesmen more than engineers, contrasting this with serious researchers like Demis Hassabis who hold a stricter definition and lament the term's reduction to a marketing tool by figures like Brockman and Altman.