Toggle light / dark theme

Monitoring blood-glucose levels and injecting insulin to keep them in a safe range is a never-ending headache for sufferers of type 1 diabetes. A number of research projects have made promising steps recently to promise easier ways of doing things, and now this type of convenience is set to move out of the lab and into the real-world. For the first time, the US Food and Drug Administration (FDA) has approved a so-called artificial pancreas designed to both monitor and inject insulin automatically, requiring minimal input from the user.

In a healthy person, beta cells in the pancreas secrete vital insulin, which in turn regulates blood-sugar levels. But for sufferers of type 1 diabetes, this process breaks down along the way, requiring them to administer finger-prick blood tests to keep tabs on their insulin levels and inject the hormone as required.

For years, scientists have been exploring better ways to keep the condition in check. These have included implanting beta cells, tracking glucose levels through contact lenses and ways insulin can be delivered via a capsule rather than a needle. But perhaps the most attractive solution is what is known as a closed-loop system, which seeks to automate both monitoring and administration of insulin to dramatically reduce the burden on the user.

Read more

Industry leaders in the world of artificial intelligence just announced the Partnership on AI. This exciting new partnership was “established to study and formulate best practices on AI technologies, to advance the public’s understanding of AI, and to serve as an open platform for discussion and engagement about AI and its influences on people and society.”

The partnership is currently co-chaired by Mustafa Suleyman with DeepMind and Eric Horvitz with Microsoft. Other leaders of the partnership include: FLI’s Science Advisory Board Member Francesca Rossi, who is also a research scientist at IBM; Ralf Herbrich with Amazon; Greg Corrado with Google; and Yann LeCun with Facebook.

Though the initial group members were announced yesterday, the collaboration anticipates increased participation, announcing in their press release that “academics, non-profits, and specialists in policy and ethics will be invited to join the Board of the organization.”

Read more

Translating from one language to another is hard, and creating a system that does it automatically is a major challenge, partly because there are just so many words, phrases and rules to deal with. Fortunately, neural networks eat big, complicated data sets for breakfast. Google has been working on a machine learning translation technique for years, and today is its official debut.

The Google Neural Machine Translation system, deployed today for Chinese-English queries, is a step up in complexity from existing methods. Here’s how things have evolved (in a nutshell).

Read more

The world’s largest technology companies hold the keys to some of the largest databases on our planet. Much like goods and coins before it, data is becoming an important currency for the modern world. The data’s value is rooted in its applications to artificial intelligence. Whichever company owns the data, effectively owns AI. Right now that means companies like Facebook, Amazon, Alphabet, IBM and Microsoft have a ton of power.

In an act of self-governance, these five companies came together today to announce the launch the new Partnership on AI. The group is tasked with conducting research and promoting best practices. Practically, this means that the group of tech companies will come together frequently to discuss advancements in artificial intelligence. The group also opens up a formal structure for communication across company lines. It’s important to remember that on a day-to-day basis, these teams are in constant competition with each other to develop the best products and services powered by machine intelligence.

Financial support will be coming from the initial tech companies that are members of the group, but in the future, membership and involvement is expected to increase. User activists, nonprofits, ethicists and other stakeholders will be joining the discussion in the coming weeks.

Read more

The lights dimmed inside the Shenzhen Bay Sports Stadium as the countdown to the match began. “Wu, si, san, er, yi!” A chime sounded and two teams of robots sprang into action across an intricately constructed battlefield. In the stands, thousands of fans cheered, and groups of small children beat red and blue balloons together, producing a percussive roar.

Each team had four rovers, nimble infantry units that quickly spread over the terrain. The rovers were shaped like small cars, but could also slide side to side, strafing like water bugs over the surface of a lake. They fired small plastic marbles from cannons mounted on top of their frames. Lumbering alongside the nimble rovers was each team’s hero, a larger tank-like robot that could fire the small plastic marbles as well as more powerful golf balls.

The heavy favorite in this matchup of RoboMasters, an annual competition held each summer, was team 1.5S, returning champions hailing from China’s University of Electronic Science and Technology in the Sichuan province. They were taking on StarPro, from the Huazhong University of Science and Technology in Wuhan.

Read more

In Brief.

  • Programmers have developed an artificial intelligence that can play the video game Doom, and do so very well.
  • Although it’s just a game, this raises questions about AI governance. How can it become more informed, integrated, effective, and anticipatory?

Read more

This video realized by the AI Lab of SoftBank Robotics shows how Pepper robot learns to play the ball-in-a-cup game (“bilboquet” in French). The movement is first demonstrated to the robot by guiding its arm.

From there, Pepper has to improve its performance through trial-and-error learning. Even though the initial demonstration does not land the ball in the cup, Pepper can still learn to play the game successfully.

The movement is represented as a so-called dynamic movement primitive and optimized using an evolutionary algorithm. Our implementation uses the freely available software library dmpbbo: https://github.com/stulp/dmpbbo.

After 100 trials, Pepper has successfully optimized its behavior and is able to repeatedly land the ball in the cup.

Read more

Komatsu’s latest autonomous truck fully embraces the notion of unmanned operation by ditching the cabin and adopting a design that optimizes load distribution and doesn’t distinguish between forwards and backwards.

Komatsu began trials of its Autonomous Haulage Systems (AHS) in a partnership with mining company Rio Tinto in 2008, and since then the technology has hauled hundreds of millions of tonnes of material in Chile and Australia’s Pilbara region.

The autonomous haul trucks like the 930E model used by Rio Tinto incorporate controls, wireless networking and obstacle detection to enable unmanned operation, but they still look like conventional mining trucks complete with driver cabins.

Read more