OpenAI and Los Alamos National Laboratory (LANL) have announced a partnership to develop evaluations for the safe use of multimodal AI models in laboratory settings, aiming to advance bioscientific research.
This collaboration continues the U.S. tradition of public-private partnerships to drive innovation in critical areas like healthcare and bioscience. The partnership responds to the recent White House Executive Order on AI safety, tasking national laboratories to evaluate frontier AI models’ capabilities, including biological applications.
OpenAI’s technology, already utilised by companies like Moderna and Color Health, aims to enhance the speed and impact of scientific research.
“We’re thrilled to announce a first-of-its-kind partnership with Los Alamos National Laboratory to study bioscience capabilities,” said Mira Murati, OpenAI’s Chief Technology Officer. “This partnership marks a natural progression in our mission, advancing scientific research while also understanding and mitigating risks.”
The joint evaluation study will assess how models like GPT-4o can assist with tasks in a physical laboratory setting through multimodal capabilities such as vision and voice. This includes biological safety evaluations for GPT-4o and its real-time voice systems, aiming to support bioscience research.
“The potential upside to growing AI capabilities is endless,” said Erick LeBrun, research scientist at Los Alamos. “However, measuring and understanding any potential dangers or misuse of advanced AI related to biological threats remain largely unexplored.
The study will build upon OpenAI’s existing work on biothreat risks and follow their Preparedness Framework, consistent with their commitments to Frontier AI Safety from the 2024 AI Seoul Summit.
The upcoming evaluation will test multimodal frontier models in a lab setting by assessing the performance of both experts and novices on standard laboratory tasks. Tasks will include genetic transformation, cell culture, and cell separation. The goal is to quantify how GPT-4o can enhance task completion and accuracy, potentially upskilling both professionals and novices in biological tasks.
LANL was established in 1943 as part of the Manhattan Project during World War II. Initially known as Project Y, it was a top-secret site dedicated to designing and developing the first atomic bomb. The laboratory is located about 35 miles northwest of Santa Fe, New Mexico.
