Facebook research creates controllable game characters from videos and more from TWiML & AI

1024 683 This Week in Machine Learning & AI

Bits & Bytes

  • Google releases MorphNet as open source. Detailed in a paper released last year, MorphNet takes an existing neural network as input and produces a new neural network that is smaller, faster and yields better performance tailored to a new problem. A new open source TensorFlow implementation is now available.
  • Facebook research creates controllable game characters from videos. The Vid2Game paper from Facebook AI Research uses a neural network to analyze videos of people performing actions, like playing sports, and then recreate that character in a scene with motion that can be controlled with a joystick. [Video]
  • IBM winds down Watson’s work in drug discovery. IBM has stopped the sale of its AI-based drug discovery and development program due to poor financial returns. IBM previously rolled out high-profile partnerships with Pfizer, Novartis, and Illumina for cancer research.
  • Facebook confirms it’s working on a voice assistant for Portal and Oculus products. Facebook has confirmed a report saying it’s working on an AI-based digital voice assistant in the vein of Amazon’s Alexa and Google Assistant.
  • Microsoft injects funding into Aussie AI-for-Earth projects. Microsoft has offered support to six Australian AI projects, taking its AI for Earth funding initiatives to US$50M around the world. The program’s participants will receive a mix of Microsoft Azure cloud computing resources and data labeling services. Check out our conversation with Lucas Joppa to hear more about Microsoft’s AI for Earth.
  • Salesforce introduces new Einstein services. Salesforce added translation and OCR to its portfolio of Einstein Platform Services, which admins and developers can easily deploy to any Salesforce page.

 

Dollars & Sense

To receive the Bits & Bytes to your inbox, subscribe to our Newsletter.

Leave a Reply

Your email address will not be published.