Power/Performance Bits: July 21

AI hardware
Researchers at Purdue University, University of California San Diego, Argonne National Laboratory, University of Louisville, Brookhaven National Laboratory, and University of Iowa developed hardware that can learn skills, offloading some of the energy needed by AI software.

“Software is taking on most of the challenges in AI. If you could incorporate intelligence into the circuit components in addition to what is happening in software, you could do things that simply cannot be done today,” said Shriram Ramanathan, a professor of materials engineering at Purdue University.

The team created artificial “tree-like” memory in a piece of potential hardware at room temperature. Researchers in the past have only been able to observe this kind of memory in hardware at temperatures that are too low for electronic devices.

The tree-like memory used in AI software organizes information into “branches,” a brain-inspired strategy that makes information easier to retrieve.

“Humans memorize things in a tree structure of categories. We memorize ‘apple’ under the category of ‘fruit’ and ‘elephant’ under the category of ‘animal,’ for example,” said Hai-Tian Zhang, a postdoctoral fellow in Purdue’s College of Engineering. “Mimicking these features in hardware is potentially interesting for brain-inspired computing.”

Researchers have developed new hardware for artificial intelligence. (Credit: Purdue University image/Qi Wang)

To create the memory, the team introduced a proton to the quantum material neodymium nickel oxide. Applying an electric pulse to the material moves around the proton. Each new position of the proton creates a different resistance state, which creates an information storage site called a memory state. Multiple electric pulses create a branch made up of memory states.

“We can build up many thousands of memory states in the material by taking advantage of quantum mechanical effects. The material stays the same. We are simply shuffling around protons,” Ramanathan said.

Through simulations of the properties discovered in this material, the team showed that the material is capable of learning the numbers 0 through 9. The ability to learn numbers is a baseline test of artificial intelligence.

“This discovery opens up new frontiers for AI that have been largely ignored because implementing this kind of intelligence into electronic hardware didn’t exist,” Ramanathan said.

“Protons also are natural information transporters in human beings,” added Zhang. “A device enabled by proton transport may be a key component for eventually achieving direct communication with organisms, such as through a brain implant.”

Recharging in motion
Engineers at Stanford University built a system that, if scaled up, could wirelessly charge objects in motion – such as electric vehicles, factory robots, or drones.

The wireless charger works like those for smartphones, creating a magnetic field that oscillates at a frequency that creates a resonating vibration in magnetic coils on the receiving device. One problem with these chargers, however, is that the resonant frequency changes if the distance between the source and receiver changes by even a small amount.

The team’s initial wireless charger, developed three years ago, was capable of overcoming this issue. It incorporated an amplifier and feedback resistor that allowed the system to automatically adjust its operating frequency as the distance between the charger and the moving object changed.

The amplifier consumed so much electricity in the initial design that the system only transmitted 10% of the power flowing through the system, which was not practical. The latest design boosts that to 92% by using a far more efficient switch-mode amplifier.

“This is a significant step toward a practical and efficient system for wirelessly re-charging automobiles and robots, even when they are moving at high speeds,” said Shanhui Fan, an electrical engineer at Stanford. “We would have to scale up the power to recharge a moving car, but I don’t think that’s a serious roadblock. For re-charging robots, we’re already within the range of practical usefulness.”

The new lab prototype can wirelessly transmit 10 watts of electricity over a distance of two or three feet. Fan says there aren’t any fundamental obstacles to scaling up a system to transmit the tens or hundreds of kilowatts that a car would need. He says the system is more than fast enough to re-supply a speeding automobile. The wireless transmission takes only a few milliseconds, which is a fraction of the time it would take a car moving at 70 miles an hour to cross a four-foot charging zone. The only limiting factor, Fan said, will be how fast the car’s batteries can absorb all the power.

Sid Assawaworrarit, a graduate student at Stanford, added that there shouldn’t be health risks associated with the charger as even ones powerful enough for cars produce magnetic fields that are well within established safety guidelines.

Even if deploying such a system for roadway use is unlikely in the near future, the researchers think that rooftop pads for drones to recharge while hovering over or charges embedded in floors for continual operation of factory or warehouse robots are feasible applications.