Robotics + Deep Learning - I think we just quietly passed a milestone where robots using deep learning can perform useful real world tasks (selecting items, packing boxes etc)

If true we could be at a watershed for robotics adoption and progress as large scale deployments generate the data to train on more complex tasks, leading to more deployments and so snowballing onwards

This seems like a much more likely process that will lead to a type of “general AI” than researchers pushing us all the way there

Covariant AI (and their partnerships) is what got me thinking: https://covariant.ai/