This team established a set of rules that govern intrinsic reward of a game-playing agent. Interestingly, their “curiosity-driven” learning model achieves similar performance when compared to conventional, extrinsically-rewarded learning models.
What happens when Siri and Wolfram Alpha fail us? We are still in need of humans to interpret and answer questions. This paper explores how expert knowledge and crowdsourcing overlap with the monetary and social benefits of popular Q&A systems.
This talk extends previous work on the YOLO algorithm to detect and classify objects in realtime. Redmon goes onto discuss the potentially nefarious ways (i.e., dual use) this technology can be used and exploited.