Bits & Bytes

AI luminary Andrew Ng launched a successor to his famous Stanford online machine learning course. The new course—actually a sequence of courses—is focused on deep learning and hosted over on Coursera. I’m planning to work through it. Let me know if you are too!
According to an SEC filing, Ng is also planning to launch a $150 million VC fund called AI Fund, L.P. It’ll be interesting to see the deals he funds via this fund.
An OpenAI bot trained using reinforcement learning beats professional Dota 2 video game players in 1v1 matches. There’s some discussion in the community as to whether the bot’s use of hardcoded rules and the Dota bot API take away from the victory. Also good discussion on the significance of the feat, .e.g. as compared to AlphaGo.
Google presented a paper at the recent KDD conference on Vizier, a system for hyperparameter optimization, i.e. model tuning.
A nice piece in the NY Times highlights the work of researchers from OpenAI, Berkeley and elsewhere on AI safety. To dig deeper into the topic, consider the materials cited in 80,000 hours’ AI safety syllabus.
MIT CSAIL researchers developed Pensieve, a system that uses reinforcement learning to optimize streaming video delivery and reduce buffering. Pensieve will be presented at next week’s SIGCOMM conference.