An energy-efficient chip called NeuRRAM fixes an old design flaw to run large-scale AI algorithms on smaller devices, reaching the same accuracy as wasteful digital computers.

By Allison Whitten

Artificial intelligence algorithms cannot keep growing at their current pace. Algorithms like deep neural networks — which are loosely inspired by the brain, with multiple layers of artificial neurons linked to each other via numerical values called weights — get bigger every year. But these days, hardware improvements are no longer keeping pace with the enormous amount of memory and processing capacity required to run these massive algorithms. Soon, the size of AI algorithms may hit a wall.

And even if we could keep scaling up hardware to meet the demands of AI, there’s another problem: running them on traditional computers wastes an enormous amount of energy. The high carbon emissions generated from running large AI algorithms is already harmful for the environment, and it will only get worse as the algorithms grow ever more gigantic.

One solution, called neuromorphic computing, takes inspiration from biological brains to create energy-efficient designs. Unfortunately, while these chips can outpace digital computers in conserving energy, they’ve lacked the computational power needed to run a sizable deep neural network. That’s made them easy for AI researchers to overlook.

That finally changed in August, when Weier Wan, H.-S. Philip Wong, Gert Cauwenberghs and their colleagues revealed a new neuromorphic chip called NeuRRAM that includes 3 million memory cells and thousands of neurons built into its hardware to run algorithms. It uses a relatively new type of memory called resistive RAM, or RRAM. Unlike previous RRAM chips, NeuRRAM is programmed to operate in an analog fashion to save more energy and space. While digital memory is binary — storing either a 1 or a 0 — analog memory cells in the NeuRRAM chip can each store multiple values along a fully continuous range. That allows the chip to store more information from massive AI algorithms in the same amount of chip space.

As a result, the new chip can perform as well as digital computers on complex AI tasks like image and speech recognition, and the authors claim it is up to 1,000 times more energy efficient, opening up the possibility for tiny chips to run increasingly complicated algorithms within small devices previously unsuitable for AI like smart watches and phones.

Researchers not involved in the work have been deeply impressed by the results. “This paper is pretty unique,” said Zhongrui Wang, a longtime RRAM researcher at the University of Hong Kong. “It makes contributions at different levels — at the device level, at the circuit architecture level, and at the algorithm level.”

By Robin Farmanfarmaian

Predictive AI applications are relatively new to mental and behavioral health, but are already showing a lot of promise. In a recent publication on detecting suicide risk through analyzing text messages, UW Medicine researchers found that algorithms performed as well as trained evaluators. This is great news for predictive AI and the ability to save lives at risk for suicide through data analysis in real-time, when and where the individual is located. This is important because some healthcare providers may be concerned when they communicate by text message with a patient, they might miss something they are trained to pick up from voice inflection, facial expression, and other auditory or physical signals. Algorithms like this can help enhance the provider’s ability to analyze the patient when communicating by text, an increasingly popular way for people to access mental health.

Beyond text messaging, there are many companies already working on analyzing a person’s speech through vocal biomarkers. Vocal biomarkers describe using someone’s voice and speech as vital signs. Digitizing the human voice and metricizing the various features of voice and speech means software programs can find patterns and detect small changes humans might not recognize. Vocal biomarker measurements and analysis for anxiety, stress, sleepiness and depression are some of the early applications.

A great example of AI voice technology that can be used directly by healthcare providers now to detect mental illness is Ellipsis Health. By harnessing the power of the human voice as a biomarker for mental health, Ellipsis Health can be used as a clinical decision support tool during clinic visits. Its technology augments the care team by helping to assess the severity of stress, anxiety, and depression.