UCF Researchers Create Brain-like Devices to Enable AI in Remote Regions and Space

Researchers at the University of Central Florida are developing new devices that will allow artificial intelligence (AI) to work from anywhere, without connecting to the internet.

This means technology ranging from natural language processing programs, like Siri or Alexa to robots and other advanced applications, could work in remote regions of the globe or even on other planets.

The researchers’ latest findings, which demonstrated a new technique to create the advanced devices, were published in a new study in the journal ACS Nano.

Currently, AI depends on connections to remote servers to perform the heavy computing and complex calculations needed to run AI processing or perform unsupervised learning, says study principal investigator Tania Roy, an assistant professor in UCF’s NanoScience Technology Center.

“Our goal is to make the artificial intelligence circuitry very small and compact,” Roy says. “That way technology like portable, handheld devices can have the circuitry on them and don’t need an internet connection. They can operate in remote areas, and have all of those functionalities, like image search or voice understanding, from any place on Earth.”

And while smartphone voice assistants are current technology that could benefit from having brain-like computing power as part of their hardware, robots are another.

“If somebody is stuck in a remote area, then the robots now will have the capacity of functioning and going to that remote area and rescuing the human being,” Roy says. “Or if we have elderly parents living alone in their homes, we can have devices that can monitor their health conditions all the time and give them some triage if something goes wrong. We would feel much more at peace if there is something to take care of them.”

For space exploration, this means robots, such as rovers, wouldn’t need a person telling them what to do.

“What happens now is that because the devices are not capable of doing unsupervised learning there is a supervisor,” Roy says. “We have to tell them what to do in the environment. But after years in space, rovers will need the power of unsupervised learning to adapt to changing environments.”

The complex, neuromorphic — or brain-like —devices the researchers have created are placed upon small, rectangular chips, about an inch wide.

Although other researchers have worked to develop this type of technology, the UCF-developed devices are more reliable due to the unique engineering and nanoscale materials they used, says the study’s lead author, Adithi Krishnaprasad ’18MS, a doctoral student in UCF’s Department of Electrical and Computer Engineering.

“We grew the material in a different way compared to how other labs grow it,” Krishnaprasad says.

“We did not grow it on some other substrate and then transfer it, rather, we grew it on the main chip itself,” she says. “We fabricated within the same platform, so that reduced the anomalies brought in by the chemistry when the transfer is used. So, we completely avoided that. By using this different technique, we have changed the way the current moves through the device. This provides better reliability by reducing variability within the device.”

The team’s advancements allow for parallelism and in-memory computing, similar to the brain, that’s required for AI and unsupervised learning, the researchers say.

The critical task of growing, or synthesizing, the nanoscale material on the chip was performed by UCF researcher Eric Jung’s group. Jung is a study co-author and an assistant professor with UCF Materials Science and Engineering, NanoScience Technology Center, and Electrical & Computer Engineering.

For their next steps, the researchers will work to further advance the technology, including building networks with the devices to enable new applications, such as image recognition.

The chips could appear in modern technology in the next 10 years, the researchers say.

Study co-authors also included Durjoy Dev ’21PhD, a graduate of UCF’s doctoral program in electrical engineering; Sang Sub Han and Changhyeon Yoo, both postdoctoral associates in the Jung Research Group at UCF; Yaqing Shen, with the Institute of Functional Nano & Soft Materials, Soochow University, Suzhou, China; Hee-Suk Chung and Tae-Sung Bae, both with the Analytical Research Division, Korea Basic Science Institute; and Mario Lanza, with the Department of Material Science and Engineering, King Abdullah University of Science and Technology in Saudi Arabia.

Roy joined UCF in 2016 and is a part of the NanoScience Technology Center with a joint appointment in the Department of Materials Science and Engineering, the Department of Electrical and Computer Engineering, and the Department of Physics. Her recent National Science Foundation CAREER award focuses on the development of devices for artificial intelligence applications. Roy was a postdoctoral scholar at the University of California, Berkeley prior to joining UCF. She received her doctorate in electrical engineering from Vanderbilt University.