New LANL machine to expand AI capabilities

Lab’s ‘Venado’ supercomputer will link research programs into data-sharing hub, officials say

LOS ALAMOS — Los Alamos National Laboratory on Monday unveiled its newest supercomputer, which officials say will accelerate how they integrate artificial intelligence into both national security work and scientific research.

The lab’s other supercomputers have been more focused on national security, whereas this system, dubbed Venado, will link the lab’s various research programs — such as climate science, astrophysics and fluid dynamics — into a centralized, data-sharing hub.

It will increase speed, efficiency and the scope of information that can be combed and pieced together. The integrative features are groundbreaking, lab Director Thom Mason said Monday during a panel discussion with government and industry leaders involved in the project.

“It’s going to allow us to solve problems that [previously] we couldn’t solve,” Mason said.

The supercomputer will serve as an institutional resource, bolstering research capabilities and workflow efficiency, he said.

The development and installation of Venado comes through a collaboration between the federal government, Hewlett Packard Enterprise and NVIDIA Corp.

“Building these systems is amazingly hard to do,” said Antonio Neri, CEO of Hewlett Packard Enterprise. “This is a marvel of engineering.”

Venado was named after one of the highest peaks in the Sangre de Cristo Mountains and was delivered to Los Alamos in March. Lab spokeswoman Laura Mullane declined to give the project’s cost, saying the lab doesn’t disclose contracting details.

Its computing capacity will draw on 2,560 Grace Hopper superchips. Hopper, a Navy rear admiral, was a pioneer in computer programming. The superchips can execute millions more instructions per second, while costing less and consuming less power, than previous chip technology, according to a news release.

The computer combines high-performing central processing and graphic processing units in a way that’s not been done before, said Jensen Huang, NVIDIA’s chief executive.

Linking the two creates a more coherent pathway that increases the flow of data and enables it to move more quickly between various programs, Huang said, which makes it quite practical.

“It can be used for everything we do,” Huang said.

Mason described Venado’s AI function as a tool that will help scientists analyze mountains of complex information. In the past, they had to reduce it to smaller, more easy-to-grasp pieces.

Supercomputing has been critical in how national labs address important problems, and Venado is expected to take problem-solving research to a new level, U.S. Deputy Energy Secretary David Turk said.

At the same time, there is concern about people being able to keep pace with AI’s rapid technological advances, including ethically, Turk said. It can do much good and at the same time much bad, he added.

Turk said public and private partnerships will not only drive progress but also create safeguards against bad actors who would use the technology with malicious intent.

Huang said the evolution of AI has been like water heating slowly until it reaches a boiling state, making technologies like Venado possible. That evolution will continue beyond what’s imagined now, he added.

“AI will revolutionize science,” Huang said.

Mason noted when supercomputers like Venado emerge, so do anxieties about machines replacing people.

“I see it as an extension of human ingenuity and not a replacement,” he said.