A View from Will Knight
Recommended Robot and AI Reads This Week
A roundup of interesting stories on robotics and artificial intelligence from other sites, collected by Will Knight, MIT Technology Review’s senior editor for AI.
Computers Learn to Create Photos of Bedrooms and Faces
A cool story about efforts to understand the way deep neural networks work. Systems previously trained to recognize faces or bedroom scenes were preprogrammed to create their own images, often revealing a twisted inner vision of the world.
Mannequins Step In for Human Billboards, but Some Are Losing Their Heads
The strange story of “robot sign-wavers,” the harassment they face, and conflicting claims to have invented the underlying sign-waving technology.
Amazon Offers New Glimpse of What Its Delivery Drone Could Look Like
The latest Amazon drone can fly like a helicopter or a plane, which is similar to the prototype unveiled some time ago by Google. Flying like a plane is far more efficient, which will be important if drones really are going to carry much more than tiny packages over any significant distance.
What It Was Like to Have a Robot Answer All of My E-mails for a Week
One journalist believes that Smart Reply, the automated e-mail response system built into Google Inbox, hints at the future of AI-mediated communication. Maybe so, but I’ve personally found its suggestions to be a bit poor so far.
Walk-Man Humanoid Robot Could Be the Future of Dangerous Work
The Italian researchers behind a humanoid robot called Walk-Man talk about their efforts to develop a machine that can take the place of a human in highly dangerous situations. This vision is some ways off, though. Earlier this year Walk-Man took part in the DARPA Robotics Challenge, and it struggled to perform even the most basic of tasks like opening doors or walking over uneven ground.
System Trains Machines to Look at Images the Way People Do
An interesting-sounding project at Microsoft Research involves fusing several different AI techniques to create something capable of parsing the contents of images in more complex and sophisticated ways.