Deep Hunt — Issue #59
Here are the highlights of this week! China creates AI anchor to read the news; In the Age of A.I., Is Seeing Still Believing?; Scaling Machine Learning at Uber with Michelangelo; Language GANs Falling Short
News
China’s state-run press agency has created an ‘AI anchor’ to read the news
Synthetic characters are slowly finding their way into mainstream culture and here’s a latest attempt by Chinese press agency — Xinhua to make AI read news. While the potential implications are many, the new host itself isn’t any more sophisticated than a CGI puppet.
Articles
In the Age of A.I., Is Seeing Still Believing?
Advances in digital imagery could deepen the fake-news crisis — or help us get out of it. In this New Yorker piece, Joshua ponders what the latest advances in AI could mean for trust on photography and videography.
Tutorials, Tools and Tips
My secret sauce to be in top 2% of a kaggle competition
Abhay Pawar from Instacart shares the tips and tricks he used to get into top 2% of Instacart’s Kaggle competition.
Scaling Machine Learning at Uber with Michelangelo
Uber built Michelangelo, our machine learning platform, in 2015. Three years later, Jeremy and Mike reflect upon the journey to scaling ML at Uber and lessons learned along the way.
Last week, we pointed to a selection of good papers presented at EMNLP. But here’s a complete summary of the happenings at EMNLP 2018 brought to you by Patrick Lewis, complete with links to the papers.
Research
This research by team at MILA has two major observations — there are fundamental flaws with quality-only evaluation. Second, by leveraging the control over quality/diversity tradeoff given by temperature param, they find MLE models constantly outperform the proposed GAN variants, over the whole quality-diversity space. You can check out the related source code here.
If you like what you are reading, please follow and recommend to your friends or give a shoutout on Twitter! I’m glad to hear your suggestions and recommendations @deephunt_in or in comments below!
