(The Center Square) The lab where early computers were used to develop the nuclear bomb is working with early AI on an evaluation study that they hope will improve artificial intelligence safety hoping to offer more intel on AI biosecurity evaluations.
“The potential upside to growing AI capabilities is endless,” Erick LeBrun, a research scientist at Los Alamos National Laboratory, said in a statement. “However, measuring and understanding any potential dangers or misuse of advanced AI related to biological threats remain largely unexplored. This work with OpenAI is an important step towards establishing a framework for evaluating current and future models, ensuring the responsible development and deployment of AI technologies.”
AI-enabled biological threats may pose substantial threats. However, previous works have never properly assessed how multimodal, frontier models might make it easier for those without advanced scientific knowledge to create a biological threat.
The team will build on past evaluations while adhering to OpenAI’s Preparedness Framework. The framework “outlines an approach to tracking, evaluating, forecasting and protecting against emerging biological risks,” a release said.
In its past evaluations, the research team concluded that ChatGPT-4 offered a modest uplift in providing information that may one day lead to the creation of biological threats. Yet, these experiments focused mostly on human performance in written tasks instead of biological benchwork. Plus, the model inputs and outputs only included text, not vision and voice data.
“Using proxy tasks and materials, the upcoming evaluation will be the first experiment to test multimodal frontier models in a lab setting by assessing experts’ abilities to perform and troubleshoot a safe protocol consisting of standard laboratory experimental tasks,” the release said.
In looking at the uplift in task completion and accuracy that ChatGPT-4 has made possible, the team wants to quantify how much frontier models can help with real-world biological tasks.
“As a private company dedicated to serving the public interest, we’re thrilled to announce a first-of-its-kind partnership with Los Alamos National Laboratory to study bioscience capabilities,” Mira Murati, OpenAI’s Chief Technology Officer, said. “This partnership marks a natural progression in our mission, advancing scientific research while also understanding and mitigating risks.”
The recently developed evaluations will support the recent White House Executive Order on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence tasks. The Department of Energy’s national laboratories are responsible for helping evaluate the capabilities of AI frontier models.
“DOE has the unique data, leadership computing, and workforce to tackle these challenges and is well suited to partner with industry on these efforts,” the release said.
Los Alamos created the AI Risks and Threat Assessments Group (AIRTAG) with a focus on developing strategies to understand the benefits, limit the risks, and help promote the deployment of AI tools safely.
“This type of cooperation is a great example of the type of work that AIRTAG is trying to foster to help understand AI risk, and ultimately making AI technology safer and more secure,” Nick Generous, deputy group leader for Information Systems and Modeling at Los Alamos, said.
