Just realized that we used to celebrate tasks that ML couldn‘t yet do because they showed us limitations of current approaches and the richness of the world. Recently it seems like we focus much more on what works, often with quite some cherry picking.