Recommended Reads on the Robot and AI Beat This Week

IHMC’s Atlas Robot Learning to Do Some Chores
One of the teams involved in the DARPA Robotics Challenge, from IHMC in Florida, has now taught its robot to do some mundane chores including grabbing objects form the floor and sweeping with a broom. The million-dollar robot, called Atlas, was developed by Boston Dynamics, a company that specializes in robots capable of dynamic, legged locomotion. Sadly, the IHMC project suggests that robot butlers are probably still a long way away.

U.S. Proposes Spending $4 Billion on Self-Driving Cars
The U.S. government is apparently keen to ensure that the U.S. maintains the lead it has established in automated-driving technologies. The Obama administration announced last week at the Detroit Auto Show that it would smooth out regulations and invest in research that could help bring the technology to market. It’s an important move. Despite so much apparent progress, significant regulatory, legal, and technical challenges remain.

Microsoft Neural Net Shows Deep Learning Can Get Way Deeper
A piece in Wired looks at the deep-learning technique used by Microsoft researchers to win an important contest that gauges the ability of AI software to recognize the contents of images. The method being used by Microsoft is one of many efforts to improve deep learning, but it’s definitely worth noting. And the piece makes an important point about efforts to automate the design of the very deep neural networks that are key to advances in AI.

Alphabet Shakes Up Its Robotics Division
In 2013, Google went on an impressive spending spree, snapping up all sorts of cutting-edge robotics companies. But since then the company’s robot program has been hit by several setbacks, most notably the loss of the executive originally behind the effort, Andy Rubin, in 2014. Alphabet, the new parent company that oversees Google and its spinouts, is now reorganizing the effort to give it greater focus.