OpenAI and Los Alamos Lab Collaborate on AI Biosecurity Research

OpenAI is partnering with Los Alamos National Laboratory to investigate the application and risks of artificial intelligence within scientific research, emphasizing biosecurity. The partnership is set to explore the potential of AI, particularly OpenAI’s GPT-4o model, to enhance lab processes while pinpointing possible dangers. The Los Alamos National Laboratory, commonly referred to as Los Alamos or LANL, is one of the United States Department of Energy’s sixteen research and development laboratories. It is renowned for its pivotal contribution to the creation of the first atomic bomb. Los Alamos has created the AI Risks and Threat Assessments Group (AIRTAG) dedicated to formulating strategies that comprehend the advantages and minimize the risks, thereby facilitating the secure implementation of AI technologies. AI in Scientific Environments Los Alamos National Laboratory will collaborate with OpenAI to assess AI’s effectiveness in research settings. They will focus on the GPT-4o model, including its voice assistant features, to see how it supports scientists. This venture is considered the inaugural study of its kind on AI biosecurity within lab environments. A key focus of this project is biosecurity. Previous studies indicated that AI models such as ChatGPT might provide knowledge that could be misused to create biological hazards. The collaboration will examine how multimodal AI could simplify the creation of these threats for non-experts, highlighting the necessity for risk mitigation strategies. Los Alamos has expressed urgency in tackling these concerns, whereas OpenAI has been more cautious in its statements. LANL writes: “AI-enabled biological threats could 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 a biological threat. The team’s work will build upon previous work and follow OpenAI’s Preparedness Framework, which outlines an approach to tracking, evaluating, forecasting and protecting against emerging biological 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. However, 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.” Evaluating AI’s Benefits and Limitations Beyond identifying risks, the partnership aims to explore AI’s advantages in research. Both entities seek to improve the efficiency and precision of scientific experiments through AI. They also intend to understand the challenges involved in deploying AI in laboratory settings. This balanced approach will provide a comprehensive evaluation of AI’s potential. The outcomes of this research could influence AI applications across various scientific disciplines. By setting a precedent for AI integration in complex scientific tasks, OpenAI and Los Alamos aim to shape future technological advancements. This partnership might redefine the landscape of scientific research, enhancing its efficiency and accuracy. The research will assess how AI can assist with real-life laboratory protocols. While AI can generate accurate protocols, executing these protocols correctly remains a challenge for non-experts. The research will examine AI’s capacity to aid individuals in learning and performing lab tasks, highlighting both its benefits and shortcomings in scientific work.