Google DeepMind researchers, including Edward Hughes and Michael Dennis, have released a new position paper emphasizing the importance of open-endedness in achieving Artificial Superhuman Intelligence (ASI). The paper, presented at the ICML 2024 conference in London, argues that the necessary components to achieve open-endedness in AI systems are now in place. Open-endedness is defined as the ability to continuously generate novel and learnable artifacts, which the researchers claim is essential for the development of ASI. Co-authors J Parker-Holder and F Behbahani also contributed to the paper, which aims to inspire rapid and safe progress towards ASI through the use of foundation models trained on internet-scale data. The paper proposes a formal definition of open-endedness to guide future research.