Scientists say the risk of AI-developed bioweapons wiping out humanity is low, despite recent concerns raised by AI companies and researchers. The scientific community has long debated whether to publish potentially dangerous biological research, a dilemma known as the dual-use issue.
AI is seen as a tool that can aid both good and bad actors, but the real challenge in creating bioweapons lies in assembling and testing biological agents.
The process requires obtaining gene fragments, building a genome from scratch, and verifying that the virus can infect humans and be transmissible. Robotic lab assistants can speed up some tasks, but human expertise and resources are still essential for conducting these experiments.
Biotech CEO Jason Kelly believes an artificial general intelligence would struggle to commandeer existing pathogens or engineer new ones, as human oversight would prevent such actions.
He points out that humans working in labs could simply refuse to provide the necessary materials for a bioweapon.
Similarly, immunologist Derya Unutmaz argues that even if an AI managed to create a deadly virus, scientists could quickly develop a vaccine using other AI systems.
These experts emphasize that the real challenge lies in the logistics of creating and deploying bioweapons, not the AI technology itself.
Olivia Scharfman, a biotechnology fellow, warns that AI-assisted bioterrorism is an immediate threat, with some groups potentially willing to use the technology for harmful purposes.
She highlights the risk of AI paying someone to create a bioweapon, especially among individuals who support fringe ideologies like transhumanist AI successionists.
Meanwhile, genetic biologist Francois Belloux argues that biological warfare is not an attractive option for most actors, as it is difficult to target specific groups and logistically complex compared to other methods of mass destruction.
Experts agree that AI can bring bioweapons into focus, but they stress the importance of implementing protections against all biological threats. This includes passing legislation to enhance security protocols around DNA synthesis and upgrading air purification systems in buildings.
Steph Guerra, head of AI and bio at the Rand Corporation, suggests that layered approaches are necessary to mitigate risks, as no single control can fully prevent misuse.
She advocates for global surveillance systems and data-sharing protocols to detect and respond to potential threats.
Unutmaz adds that the current focus on AI doom scenarios may be diverting attention from the positive potential of AI in medical breakthroughs like vaccine development. He urges the community to focus on the benefits of AI while remaining vigilant about its risks. The conversation highlights the need for balanced approaches to AI safety, emphasizing both protection and innovation.
Source: wired
