The faster things go now, the more likely it is to be a slow takeoff scenario. AI models, though they are getting increasingly close to matching human performance on general tasks, are still very far from matching human parameter count in any efficiency scale close to the human brain. This will be a requirement before general ASI can bring about an intelligence explosion, which I still don't see happening before 2040. Meanwhile I believe we are in the midst of a slow takeoff now that will usher in enormous societal change with proto-agi and agi and agi systems.