The International Conference on Machine Learning, ICML, is one of the three flagship machine-learning conferences alongside NeurIPS and ICLR, and its own about page calls it “one of the fastest growing artificial intelligence conferences in the world.” Its roots go back to an International Workshop on Machine Learning first held at Carnegie Mellon University in July 1980, organized by Jaime Carbonell, Ryszard Michalski, and Tom Mitchell.
That first gathering reflected the machine learning of its era, much of it symbolic and rule-based rather than the neural-network methods that now dominate. The workshop ran roughly annually through the 1980s, and in 1993 the series formally became the International Conference on Machine Learning. It is today organized by the International Machine Learning Society.
ICML’s long arc captures the shifting center of the field. Its early proceedings document decision-tree learning, inductive logic, and reinforcement learning; by the 2010s the same conference had become a primary venue for deep learning, optimization, and large-scale statistical methods, tracking the discipline’s move from hand-built rules toward learning from data at scale.
Why business readers should care: ICML’s forty-plus-year history shows that machine learning is a mature research discipline with deep foundations, not a sudden 2020s invention, and its proceedings remain a leading indicator of which methods will reach products next.