News

OpenAI, Los Alamos Lab Study Potential of AI Misuse for Bio Threats

• By John K. Waters
• 07/10/2024

Researchers at the Los Alamos National Laboratory (LANL) announced this week a new collaboration with generative AI pioneer OpenAI on an evaluation study of how multimodal, frontier models could lower the barrier to entry for non-experts seeking to create biological threats.

"The potential upside to growing AI capabilities is endless," said Erick LeBrun, research scientist at Los Alamos, 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 have the potential to pose a significant risk, but existing work has not assessed how multimodal, frontier models could lower the barrier of entry for non-experts to create biological threats. A "frontier model" in the context of artificial intelligence refers to the most advanced and cutting-edge AI models that push the boundaries of current technology. These models are typically characterized by their complexity, scale, and the novel capabilities they bring to a range of applications. Frontier models often include the latest advancements in machine learning, deep learning, natural language processing, computer vision, and multimodal capabilities (the ability to process and integrate multiple forms of data such as text, images, and audio).

The team's work will build on previous work and follow OpenAI's Preparedness Framework, which outlines an approach to tracking, evaluating, forecasting, and protecting against emerging biological risks. 

This partnership follows a long tradition of the U.S. public sector, and in particular the national labs, working with the U.S. private sector to ensure advances in innovation translate to advancements in essential areas, such as health care and bioscience, the companies said.

"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," said Mira Murati, OpenAI's Chief Technology Officer, in a statement. "This partnership marks a natural progression in our mission, advancing scientific research, while also understanding and mitigating risks." 

In previous evaluations, the research team found that ChatGPT-4 provided "a mild uplift" in providing information that could lead to the creation of biological threats. But these experiments focused on human performance in written tasks, rather than biological benchwork, and model inputs and outputs were limited to text, which excluded vision and voice data. 

Because the researchers will be 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. Examining the uplift in task completion and accuracy enabled by the latest version of ChatGPT (ChatGPT-4o), will allow the team to quantify and assess how frontier models can assist in real-world biological tasks. 

The announcement of this partnership is timely, following the recent White House Executive Order on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence. This directive emphasizes the role of the Department of Energy's national labs in evaluating advanced AI models, including those with biological capabilities. OpenAI's collaboration with LANL aligns with this national priority, the companies say, aiming to harness the potential of AI to amplify scientific discovery for the greater good.

OpenAI's technology is already making inroads into the healthcare sector. For example, Moderna is using AI to streamline clinical trials, and Color Health employs GPT-4o to aid healthcare providers in making informed decisions about cancer treatment. This new partnership aims to extend such innovations into bioscience laboratories, exploring how AI can assist scientists in conducting complex biological tasks.

Among other things, this study will assess how AI can enhance the performance of both experts and novices in laboratory tasks, from genetic transformations to cell culture and separation. By quantifying the impact of AI on task completion and accuracy, OpenAI and LANL aim to set new standards for AI applications in bioscience.