Mustafa Suleyman, co-founder of Google DeepMind and an AI expert, has raised concerns about the potential misuse of artificial intelligence to create synthetic viruses that could spark pandemics. Speaking on a podcast, Suleyman warned that individuals might experiment with engineering pathogens to be more transmissible or lethal, either accidentally or intentionally. To prevent such scenarios, he called for restrictions on access to advanced AI technology and software that can be used for this purpose.

Suleyman emphasized the need for containment measures, similar to existing restrictions on accessing pathogenic microbes like anthrax. He underscored the importance of limiting who can use AI software, cloud systems, and biological material. Additionally, he suggested restricting access to certain substances. He argued for a precautionary approach when developing AI, where potential risks are considered and precautions are taken.

Suleyman’s concerns align with a recent study that highlighted how even individuals without a biology background, such as undergraduate students, can obtain suggestions for bio-weapons through AI systems. The study found that chatbots could identify potential pandemic pathogens and provide detailed protocols for their generation from synthetic DNA. It warned that large language models like ChatGPT could make pandemic-class agents accessible to people with minimal laboratory training.

The researchers behind the study proposed non-proliferation measures to address this issue. These measures include third-party evaluations of language models before their release, curating training datasets to remove harmful concepts, and implementing verifiable screening of DNA synthesis orders and usage in organizations involved in genetic engineering.

Ensuring the responsible use of AI in biological research and restricting access to potentially harmful technologies emerges as an urgent concern. As AI continues to advance, it will be crucial to prioritize safety and mitigate any risks associated with its applications.

Sources:
– “Synthetic viruses generated through AI ‘could be used to spark pandemics’,” The Independent
– “Language models as threats to biosecurity,” PLOS ONE