Amazon's titchy robots hit the streets, Waymo starts a self-driving car factory...

Roundup Hello, here’s this week’s roundup of all things robot and AI-related. We’re talking delivery robots, self-driving cars, and a new natural language dataset to play with.

Autonomous delivery robots are coming: Amazon is testing out an autonomous delivery service using a small, six-wheeled robot called Scout.

Scout relies on electrical power and preprogrammed instructions to autonomously deliver Amazon packages to their rightful owners. The service is currently being rolled out in a neighborhood in Snohomish County, Washington State.

A small fleet of six robots will work from Monday to Friday during the day. Amazon said it had it been developed by its R&D lab in Seattle, and that it can safely worm its way around pedestrians, pets and other obstacles. Scout looks like a cooler on wheels, and is designed to trundle along the pavement at walking speed.

Although it can, apparently, navigate to its destination on its own it will be supervised by an Amazon employee during its trips to and from houses. Hopefully, a human presence will probably deter people from kicking the box on wheels or stealing its contents.

We need Waymo cars: Waymo is building a factory in Michigan to manufacture and test its self driving cars.

The Michigan Economic Development Corporation, an agency supporting job growth, approved Waymo’s plans. Michigan is known for its once-bustling automotive industry as well as its cold winters, making it an ideal environment to build and test Waymo’s cars in tough conditions.

“Waymo will first identify a facility in Southeast Michigan and, over the next few years, aim to create hundreds of local jobs in the community,” it said.

"We’ll be looking for engineers, operations experts, and fleet coordinators to join our team and help assemble and deploy our self-driving cars. This will be the world’s first factory 100%-dedicated to the mass production of L4 autonomous vehicles.

Employees will work to integrate Waymo’s hardware and software systems onto its fleet of Fiat Chrysler and Jaguar Land Rover vehicles.

Apple’s self driving car project is downsizing: Menawhile, two hundred jobs are being slashed at Apple’s self-driving car project.

Apple has been tight-lipped about Project Titan. It has been quietly working on autonomous vehicles for a while, and applied for an permit to test its cars on the roads of California back in 2017. Researchers also published a paper describing a neural network system, dubbed Voxelnet, capable of detecting obstacles like pedestrians and cyclists.

There was a bit of reshuffling as Doug Field, an engineer currently leading Project Titan, was wooed back to Apple last year after he left to become Tesla’s VP of engineering. But it looks like things haven’t been going exactly swimmingly since.

“We have an incredibly talented team working on autonomous systems and associated technologies at AppleAn Apple spokesperson told CNBC.

"As the team focuses their work on several key areas for 2019, some groups are being moved to projects in other parts of the company, where they will support machine learning and other initiatives, across all of Apple.”

Google’s Q&A dataset: The Chocolate Factory has released a new dataset to test natural language processing systems.

Dubbed Natural Questions, a dataset containing 300,000 questions along with corresponding answers scraped from Wikipedia pages. AI nerds have been using question and answer tasks to test a machine’s understanding of text and images.

Machine reading is seen as a path towards computers developing some sort of common sense and reasoning skills. If computers could understand and communicate using text, well, it would make life a whole lot easier and interesting.

At the moment, AI systems can be pretty good at answering general knowledge questions by matching up relevant keywords that appear in both the question and answer to spit out the correct response. It struggles more, however, with following basic events in storylines and reasoning tasks.

If you want to train your own QA system with Google’s dataset, click here.

DeepMind’s StarCraft bot: Just in case you missed the demo of DeepMind’s StarCraft II bot, AlphaStar, playing against two professional players this week, we wrote about it in depth here.

And now here’s the full replay of the event. ®

Youtube Video