Victoria Krakovna has a fantastic list of examples of AI systems that have misbehaved and subverted their way to meeting their incentive goals by sometime hilarious means! Some game systems evolved to clap body parts together to exploit a collision detection bug and generate free game enerrgy.
China's state run media unveiled the world's first AI news anchor this week. The AI anchor is generated from composites of a real Chinese news anchor with added synthesised speech.
Amit Shekhar has a short but helpful medium post on MindOrks with an example of how to use Tensorflow Lite on Android for object detection, including a link to the sample application.
The Adobe MAX 2018 event demonstrated a plethora of AI enabled "multimedia" tools from Adobe this year. The innovations include the Moving Stills project which automatically generates an animated fly-through from a static image, ProjectKazoo which allows the user to hum a tune and generate sampled instrumental music from it and Fontphoria which automatically generates a custom font from one letter of a newly designed font. Check out these and the rest of the projects on the Adobe Blog.
Google have launched a search engine for data sets similar to Google Scholar, allowing data scientists, researchers and others. The search engine presents summaries of each dataset along with links to the source location of the dataset.
Researchers at NVIDIA have used a trained deep learning model to take regular 30 frames per second (fps) video and generate from it smooth slow motion 240 fps or 480 fps video!
The Facebook Ads team have launched a field guide to machine learning video series. The six videos include work practices and tips at applying machine learning to real world problems.
Google have launched The Lever, a blog that presents the best practices in applying machine learning, including posts about how to approach adding machine learning to your product, and posts about how to gather or acquire data.
This Smashing Magazine tutorial demonstrates how to build a machine learning model that can predict which room you are in based on the signal strength of wifi networks around you. The system requires a map of the networks to be recorded and a model trained using the recorded data. Such a system could be useful to passively trigger IOT devices based on which room your phone is in, without the need for active notification methods like bluetooth beacons.
