科学家首次利用人工智能模型Evo成功创造了新型病毒[1]。这项研究由Arc Institute研究机构和Stanford University研究团队在加州帕洛阿尔托完成，相关成果已发表在《科学》杂志上[1]。研究人员通过训练AI模型识别DNA结构模式来生成病毒基因序列，该模型在训练过程中扫描了约9万亿个核苷酸[1]。
研究团队从700,000个潜在病毒序列中创造了285个DNA分子，其中16种成功存活[1]。这些新创造的病毒与天然病毒Phi X-174相似，仅能感染细菌，对人类不构成威胁[1]。为了防止危险应用，研究人员在训练模型时未向其提供任何感染人类的病毒数据[1]。
这一突破为医学进展带来了新的可能性，但同时也引发了人们对AI技术可能被用于创造危险病原体的担忧[1]。美国国立卫生研究院随后推出了新政策，禁止高风险生命科学研究[1]。
Researchers have successfully used an artificial intelligence model called Evo to create new viruses for the first time, marking a significant scientific breakthrough that has already drawn both enthusiasm and caution regarding potential misuse.[1] The study, published in Science magazine and conducted by scientists at the Arc Institute in Palo Alto, California, demonstrates the capability of AI to generate functional biological sequences by pattern recognition.[1]
The research team trained the Evo model on approximately 9 trillion nucleotides to identify DNA structure patterns and generate viral gene sequences.[1] From 700,000 potential viral sequences, researchers synthesized 285 DNA molecules, resulting in the successful creation of 16 viable new viruses.[1] All of the generated viruses resembled the natural bacteriophage Phi X-174, infecting only bacteria and posing no threat to humans.[1] To mitigate the risk of dangerous applications, the researchers deliberately withheld data on human-infecting viruses from the model during training.[1]
The breakthrough arrives amid heightened biosecurity concerns, as the U.S. National Institutes of Health unveiled new policies last week restricting high-risk life sciences research.[1] While the findings offer promise for medical advances, they have simultaneously raised concerns about the potential for artificial intelligence to be weaponized in creating hazardous pathogens.[1]