DOE laboratories showcase deep bench in artificial intelligence tech

Ten national laboratories, including Sandia, shared one of the largest booths at the AI Expo for National Competitiveness from May 7-8 in Washington, D.C. The new conference provided the DOE with an opportunity to demonstrate how the nation’s scientists, researchers and engineers are working together to advance and apply the power of artificial intelligence.

Because of recent advances such as ChatGPT, the greater population can now easily interact with and use AI, but AI is not new to Sandia. Scientists and engineers at the Labs have been creating and employing AI and machine learning algorithms for many years. Adoption of AI is rapidly accelerating due to the availability of user-centered large language models and commercial tools that can be leveraged across all fields to expedite projects and results.

More than 13,000 people registered for the AI Expo, including government stakeholders, academia and industry partners in attendance.

Jen Gaudioso, Sandia director for computing research, understood the criticality of the event. “The AI Expo provided a forum for us to exhibit some of the latest technological breakthroughs being developed for science, security and energy applications and also discuss their implications for the U.S. and allied competitiveness.”

Fifteen Sandians from six divisions attended the expo and presented state-of-the-art advancements in applied machine intelligence and virtual reality, including two demos led by Danny Gomez, Demitri Maestas, Kevin Potter and Aniket Pant.

One AI-powered demo featured in the booth has virtual reality understanding. JARVIS is an immersive extended reality environment that encouraged conference attendees to engage in a dialogue with the system. Participants were able to ask questions about the visual items they encountered within the virtual environment, and JARVIS responded with real-time, audible answers in natural language. This interactive experience showcased JARVIS’ ability to communicate in natural language and provide timely responses relevant to the virtual environment, creating a better and more enriched virtual experience.

The Machine-Assisted Anomaly Detection System demo applied machine learning to identify visual anomalies. In short, it helped users to answer more than the question, “Which one of these things is not like the others?” The detection system can locate defects among parts without having previously seen a defect example.

New Mexico Sen. Martin Heinrich, who has a background in engineering, attended the conference and stopped by the DOE booth to experience Sandia’s demos. He saw firsthand how powerful the tools could be in expediting processes that would otherwise take staff numerous hours to complete.

“New Mexico’s national labs are leading the nation in capitalizing on AI’s potential, and I was thrilled to see Sandia Labs display this up close with an immersive virtual reality experience that will be a key tool for the next generation of the engineering workforce. I will continue doing everything in my power to put our state’s national labs front and center in the development and deployment of this technology,” said Heinrich, co-founder and co-chair of the Senate AI Caucus. “The actions we take now in Congress will determine how rapid advancements in AI impact our society and country. For our national security, for our democracy and for the safety and well-being of the American people, it’s vital we get it right.”

“By utilizing AI to exploit our extensive knowledgebase of information and models, we are accelerating the rapid deployment of mission applications for national security,” said Dan Turner, who leads the AI for Nuclear Deterrence initiative at Sandia.

DOE agrees. AI is a central tenet of Frontiers of AI for Science, Security and Technology, a DOE vision seeking to dramatically accelerate the pace of research and development and enable scientific capabilities previously thought to be impossible. It also pinpoints AI as a strategically essential technology intimately tied to the nation’s future.

The national labs are aligned with the Frontiers of AI for Science, Security and Technology vision. They are jointly creating hubs using exascale computers to address DOE mission grand challenges, putting together teams to organize data for AI training and evaluation, creating tools and approaches to understand and manage AI risks, and developing a common AI platform and dedicated AI hardware resources.

Sandia’s John Feddema, senior manager for enhanced decision-making, said, “Sandia is one of the national labs involved in compiling ‘AI Killer Apps.’” According to him, these apps assist the country in monitoring for threats and facilitating deterrence, revolutionizing energy applications, augmenting next generation systems, enhancing decision making through assimilated data and making bioweapons obsolete. “AI is a powerful tool that will help solve the complex challenges we face and unlock new opportunities,” he said.

DOE headquarters noted other positive outcomes in an AI Expo event recap stating, “We have heard from many people (including DOE Deputy Secretary David Turk, DOE Under Secretary for Science and Innovation Secretary Geri Richmond and NNSA Administrator Jill Hruby) that DOE made quite the statement, and now, it’s on us to capitalize on the momentum.”

Sandia Labs Director James Peery said, “In the past 100 years, the world has realized breakthroughs in game-changing technology that have transformed civilization as we know it. The advent and global availability of AI is one of these. Wielding it with integrity for the good of our country is paramount, and there is no time to waste.”

Informing high-consequence decisions through machine learning

Trust is everything, according to Erin C.S. Acquesta. Her work on the Predictive Capability Maturity Model framework helps Sandia gather credibility evidence, identify gaps in modeling and simulation capabilities, and then communicate those to decision-makers so they can do a thorough evaluation before making decisions.

Erin presented on the topic of AI security at the expo. “Credibility is essential when using computational simulation models to inform high-consequence decisions,” she said, noting that those predictions ultimately inform engineering designs and modifications on national security applications.

“The NNSA labs have prioritized credibility of computational simulation modeling for the last 30 years, and Sandia has been leading the prioritization of defining credibility of machine learning tools used for high-consequence applications.

“With so much emphasis on AI, we need to make sure we keep the conversation going about how we strike a balance between leveraging the advantages of machine learning while ensuring its responsible use for national security purposes.”