Evaluating AI for Biosecurity: A Groundbreaking Partnership
Evaluating AI for Biosecurity: A Groundbreaking Partnership
Researchers at Los Alamos National Laboratory are teaming up with OpenAI to enhance AI safety, particularly in biosecurity. This pioneering evaluation will contribute to advanced AI biosecurity assessments.
“The potential of AI is vast,” said Erick LeBrun, a research scientist at Los Alamos. “However, understanding and mitigating the risks associated with advanced AI, especially biological threats, is crucial.”
Addressing AI-Enabled Biological Threats
AI-enabled biological threats pose significant risks. This collaboration will build on previous research and follow OpenAI’s Preparedness Framework for tracking, evaluating, and protecting against emerging biological risks. Previous evaluations showed ChatGPT-4 provided some information that could lead to biological threats but was limited to text-based inputs and outputs. The new study will test multimodal models like ChatGPT-4o in lab settings. This partnership also signals OpenAI’s potential shift toward less private sector involvement, highlighted by Apple and Microsoft relinquishing their OpenAI board seats.
Quantifying AI’s Impact on Real-World Tasks
Using proxy tasks, the evaluation will assess how models like ChatGPT-4o assist in real-world biological tasks, focusing on task completion and accuracy. This partnership supports the White House Executive Order on AI development and involves DOE national laboratories. The collaboration further suggests OpenAI’s broader engagement with public sector responsibilities, reflecting its evolving mission.
Los Alamos has also established the AI Risks and Threat Assessments Group (AIRTAG) to understand the benefits and mitigate risks in AI tool deployment. “This cooperation exemplifies AIRTAG’s mission to make AI safer and more secure,” said Nick Generous, deputy group leader at Los Alamos.
Bottom Line – AI for Biosecurity
The partnership between Los Alamos and OpenAI is a significant step in AI safety and biosecurity (AI for biosecurity), focusing on evaluating and mitigating risks associated with advanced AI models in biological contexts.
FREE AI TOOL
Free Download – Aragon Research’s New Adopting AI Assistants Checklist
AI is here but is your business ready?
Aragon’s new checklist is your roadmap to successfully integrating AI Assistants and Generative AI into your business operations.
Download Aragon’s Free AI Insights
UPCOMING AI WEBINAR
AI is Here: Increase Value & Reduce Risk with AI Governance
Almost every week, we hear about another organization facing an increasingly complex myriad of ethical questions, from AI. However, leaders don’t have either the training or the focus to be able to address these issues.
In this webinar, we explore how your organization can reduce the risk of AI with effective AI governance, and provide a checklist for how to approach defining AI governance.
- What challenges do organizations face in supporting AI?
- How can governance reduce these risks and challenges?
- How can organizations use Aragon Research’s AI Governance checklist?
Have a Comment on this?