Bits & Bytes – 8.10.17

1024 1024 The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence)

Bits & Bytes

  • Microsoft made headlines last week when it formally added AI to its corporate vision statement, dropping references to “mobile-first” “Our strategic vision is to compete and grow by building best-in-class platforms and productivity services for an intelligent cloud and an intelligent edge infused with AI.”
  • BuzzFeed trains a random forest ML algorithm to identify instances of spy planes in flight tracking data in “BuzzFeed News Trained A Computer To Search For Hidden Spy Planes. This Is What We Found.” Very interesting read, despite the characteristically clickbaity title.
  • The FastText team at Facebook AI Research release a new set of pre-trained word vectors trained on Wikipedia, news and web crawl data.
  • Facebook also announced that they were transitioning entirely to neural networks for language translation, from a phrase-based statistical system. The new system uses sequence-to-sequence LSTMs with attention. Stay tuned for a deep dive into LSTMs coming later this month!
  • OpenAI released RL-Teacher, an open-source interface for training reinforcement learning based AIs via occasional human feedback rather than mathematically expressed reward functions.
  • As a New Yorker, I consider sarcasm to be somewhat of an art form. A group of researchers from MIT Media Lab and elsewhere published a paper on DeepMoji [PDF], a deep learning model that can detect sarcasm, sentiment, and emotion from emoji.

Sign up for our Newsletter to receive the Bits & Bytes weekly to your inbox.

Leave a Reply

Your email address will not be published.