Toggle light / dark theme

Researchers make transformational AI seem ‘unremarkable’

Physicians making life-and-death decisions about organ transplants, cancer treatments or heart surgeries typically don’t give much thought to how artificial intelligence might help them. And that’s how researchers at Carnegie Mellon University say clinical AI tools should be designed—so doctors don’t need to think about them.

A surgeon might never feel the need to ask an AI for advice, much less allow it to make a for them, said John Zimmerman, the Tang Family Professor of Artificial Intelligence and Human-Computer Interaction in CMU’s Human-Computer Interaction Institute (HCII). But an AI might guide decisions if it were embedded in the decision-making routines already used by the clinical team, providing AI-generated predictions and evaluations as part of the overall mix of information.

Zimmerman and his colleagues call this approach “Unremarkable AI.”

Read more

A multi-scale body-part mask guided attention network for person re-identification

Person re-identification entails the automated identification of the same person in multiple images from different cameras and with different backgrounds, angles or positions. Despite recent advances in the field of artificial intelligence (AI), person re-identification remains a highly challenging task, particularly due to the many variations in a person’s pose, as well as other differences associated with lighting, occlusion, misalignment and background clutter.

Researchers at the Suning R&D Center in the U.S. have recently developed a new technique for person re-identification based on a multi-scale body-part mask guided attention network (MMGA). Their paper, pre-published on arXiv, will be presented during the 2019 CVPR Workshop spotlight presentation in June.

“Person re-identification is becoming a more and more important task due to its wide range of potential applications, such as , and image retrieval,” Honglong Cai, one of the researchers who carried out the study, told TechXplore. “However, it remains a challenging task, due to occlusion, misalignment, variation of poses and background clutter. In our recent study, our team tried to develop a method to overcome these challenges.”

Read more

Move over, silicon switches: There’s a new way to compute

Logic and memory devices, such as the hard drives in computers, now use nanomagnetic mechanisms to store and manipulate information. Unlike silicon transistors, which have fundamental efficiency limitations, they require no energy to maintain their magnetic state: Energy is needed only for reading and writing information.

One method of controlling magnetism uses that transports spin to write information, but this usually involves flowing charge. Because this generates heat and , the costs can be enormous, particularly in the case of large server farms or in applications like artificial intelligence, which require massive amounts of memory. Spin, however, can be transported without a charge with the use of a topological insulator—a material whose interior is insulating but that can support the flow of electrons on its surface.

In a newly published Physical Review Applied paper, researchers from New York University introduce a voltage-controlled topological spin switch (vTOPSS) that requires only electric fields, rather than currents, to switch between two Boolean logic states, greatly reducing the heat generated and energy used. The team is comprised of Shaloo Rakheja, an assistant professor of electrical and at the NYU Tandon School of Engineering, and Andrew D. Kent, an NYU professor of physics and director of the University’s Center for Quantum Phenomena, along Michael E. Flatté, a professor at the University of Iowa.

Read more

A machine has figured out Rubik’s Cube all by itself

Working strategy, by starting at the desired end destination and then looking back, by connecting the dots as they are presented chronologically (in our present) towards the future, a strategic level of thinking now available to machines.


Unlike chess moves, changes to a Rubik’s Cube are hard to evaluate, which is why deep-learning machines haven’t been able to solve the puzzle on their own. Until now.

Read more

Smarter training of neural networks

These days, nearly all the artificial intelligence-based products in our lives rely on “deep neural networks” that automatically learn to process labeled data.

For most organizations and individuals, though, deep learning is tough to break into. To learn well, neural networks normally have to be quite large and need massive datasets. This training process usually requires multiple days of training and expensive graphics processing units (GPUs)—and sometimes even custom-designed hardware.

But what if they don’t actually have to be all that big, after all?

Read more

See Tesla’s Enhanced Summon Pick up a Driver in a Parking Lot

After its release, Tesla owners could instruct their vehicles to autonomously pull in or out of a parking space or garage with the push of a button. They just couldn’t expect the car to make any turns.

In late 2018, Musk began teasing a major update to Summon, which Tesla began rolling out in March — and a newly released video of Enhanced Summon in action shows just how far autonomous tech has come in three years.

Read more

Ducati Is Working on a Futuristic Electric Motorcycle

“The future is electric,” Ducati CEO Claudio Domenicali said during an event in Spain, according to Electrek’s translation, and that the company is “not far from starting series production.”

READ MORE: Ducati CEO confirms ‘The future is electric’, says electric Ducati is coming [Electrek]

More on the bike: BMW’s Self-Driving Motorcycle Could Help Keep Bikers Safe.

Read more

/* */