Energy Exploring AI in Clean Energy, Supercomputing

The Energy Department is exploring artificial intelligence for clean energy initiatives and scientific research as officials consider responsible and secure use of the technology.

In a partnership with the Pacific Northwest National Laboratory, the agency employed AI to discover new battery materials for utilities to build and store clean energy by 2050 and have 100% clean electricity on the grid by 2035.

“We used AI that the lab and Microsoft started to identify as next-generation battery materials,” said Energy Secretary Jennifer Granholm at the NVIDIA AI Summit last month. “We started with 32 million potential materials and dwindled it down to 23 within a matter of weeks that would have taken years to be able to do, and so the acceleration of that discovery by AI is almost magical.”

Another initiative the agency is exploring is enabling AI to build a whole new generation of supercomputers, which are already housed at national labs around the country such as Oak Ridge National Laboratory in Tennessee and at Argonne National Laboratory in Chicago.

“We make these supercomputers available to researchers and companies to be able to use and we have AI embedded throughout those. We have AI test beds that support the next generation of AI accelerators. We make those all available again to companies and researchers to be able to do that next generation of work,” said Granholm.

“There’s all the talk about how dangerous AI can be and there’s all this talk about how great AI can be,” said Granholm. “We need to use AI for both offense and defense, offense to solve these big problems and defense to make sure the bad guys are not using AI for nefarious purposes or at least that we can detect and avert. Let’s just make sure it’s shaped to serve the public interest.”

During the NVIDIA event, Helena Fu, Director, Office of Critical and Emerging Technologies, at DOE, highlighted what the agency wants to do with AI applications. Instead of incremental science, the agency wants to see big step changes in capabilities.

“That ability for these models to do scientific reasoning will only bring you to the door of DOE and the national labs, but to actually advance AI systems and AI reasoning you will need access to scientific systems and our scientific systems will need access to AI systems,” said Fu. “So where we see fast here is the ability to kind of bridge these two places, both scientific reasoning and the capabilities and the new kinds of models we’ll need to develop with AI.”

What’s Ahead for AI in Security

President Biden’s AI executive order tasked the Cybersecurity and Infrastructure Security Agency (CISA) with working across the 16 critical infrastructure sectors, including the Energy Department, to assess how advances in AI would make them more vulnerable to critical failures and cyber attacks.

CISA Chief AI Officer Lisa Einstein said the agency organized assessments around three core risks: attacks using AI, attacks on AI systems and failures in design and implementation. CISA is currently conducting its second round of risk assessments that will be completed by January 2025.

“The important thing here is that we’re building a muscle memory as a community for how to collaborate and assess these risks as they advance. We have developed a group called Joint Cyber Defense Collaborative, which has brought in new AI security providers and AI companies also with other critical infrastructure owners and operations,” said Einstein. “So far this year we’ve had two tabletop exercises about AI security incidents, and it’s meant to build this enduring operational community where people can lower the barrier for information sharing, help each other adapt to new risks and threats, and we’re working on AI security incident collaboration playbook that will be published this fall.”