116
AI’s recursive self-improvement might not come so quickly after all
(www.technologyreview.com)
This is a most excellent place for technology news and articles.
They would need to be trained for open ended creative tasks, which is just hard in the current reinforcement learning paradigm.