Artificial Intelligence and Biological Misuse: Differentiating Risks of Language Models and Biological Design Tools

This paper proposes two categories of AI models that could have impacts on biosecurity in different ways: large language models (LLMs) and biological design tools (BDTs). Overall, the author suggests that LLMs could lower barriers to biological misuse, while BDTs could expand the capabilities of sophisticated actors to create pandemic pathogens worse than anything seen to date. The paper also provides suggestions for mitigating these risks.

Original text:
https://arxiv.org/abs/2306.13952

Author(s):
Jonas B. Sandbrink

A podcast by BlueDot Impact.

Chapters

• 0:16 Abstract
• 1:51 Introduction
• 3:58 Risks from large language models (LLMs)
• 12:09 Risks from biological design tools (BDTs)
• 16:48 Takeaways for risk mitigation
• 21:10 Conclusion