https://www.quantamagazine.org/machine-learnings-amazing-ability-to-predict-chaos-20180418/ These are a few of my faaaavorite things
https://research.googleblog.com/2018/04/introducing-semantic-experiences-with.html Cautiously optimistic! Lotsa stuff has promised to do this and not-quite-done-it, but hey, nothing happens until it does.
https://medium.com/activewizards-machine-learning-company/a-comparative-analysis-of-chatbots-apis-f9d240263e1d Kinda kooky how mature this tech is getting
https://medium.com/scribd-data-science-engineering/multi-armed-bandits-for-the-win-240b71bc3464 Ah, multi-armed bandits. The old "explore vs exploit" tradeoff. Or as I like to call it, "I wish I spent more time reading through that Distributed Systems book instead of browsing Reddit, but then again I found that book on Reddit to begin with."
https://blog.openai.com/evolved-policy-gradients/ Awesome stuff! Making any progress whatsoever in the quest to get ML algos to be able to generalize from experience to new contexts is awesome. Relevant:
(maybe I should do a longer post later on this topic?)