Virologist Opinions: An Important Component for the Governance of the Convergence of Artificial Intelligence and Dual-Use Research of Concern

TL;DR

This paper explores virologists' views on governing AI in life sciences, emphasizing the need for expert input in policy development.

Contribution

It highlights the importance of incorporating virologist opinions into AI governance frameworks for dual-use research.

Findings

Virologists believe AI's benefits and risks in virology are yet to be fully realized.

Experts stress the need for governance processes that include virologist input.

AI tool capabilities are hard to predict without experimental validation.

Abstract

The convergence of artificial intelligence (AI) and the life sciences has brought in silico research into policy conversations around dual-use research of concern and pathogens with enhanced pandemic potential research. This study considers the expert opinions of virologists on governance of AI and life sciences research.

Semi-structured interviews with virologists were conducted and qualitatively analyzed to explore expert opinions about AI and virology. Interviewees were asked about the risks and benefits of AI, policy development considerations, and about evaluating the capability of AI tools in the field of virology.

Interviewed virologists generally expressed similar sentiments in responses to questions, including that benefits and risks of AI use in virology research are still to come, that policy and governance should be a process that includes virologist input, and that it is challenging to predict the capability of AI tools without experimental wet-lab validation.

Governance should be informed by expert opinions of practitioners, and it is important to consider how such opinions are incorporated. Expert opinions are valuable in understanding the impact of governance measures on beneficial research and development, and ensuring that governance measures are practicable and applicable. Virologists interviewed generally had similar opinions around AI and virology topics and often expressed an expectation that their opinions would develop over time.

Given the uncertainty around the capability of AI technologies in the life sciences, it may be better to focus on developing frameworks for how governance measures will be developed, and to monitor developments, than to focus on specific interventions.