What do researchers from Anthropic, NCBI, the Broad Institute, and the Chan Zuckerberg Initiative reveal about why AI agents fail at retrievThe gap between AI and biology is not a failure of intelligence but of infrastructure — a lesson made clear by new research from Anthropic and leading scientific institutions.
AI 提示
Create a landscape editorial hero image for this Studio Global article: What do researchers from Anthropic, NCBI, the Broad Institute, and the Chan Zuckerberg Initiative reveal about why AI agents fail at retriev. Article summary: In a collaboration between Anthropic, NCBI, the Broad Institute, and the Chan Zuckerberg Initiative (CZI), researchers demonstrated that state-of-the-art AI agents fail at retrieving biological data from public databases. Topic tags: general, government, academic, general web, user generated. Reference image context from search candidates: Reference image 1: visual subject "# Artificial Intelligence agents for biological research: a survey. A **.gov** website belongs to an official government organization in the United States. Inclusion in an NLM data" source context "Artificial Intelligence agents for biological research: a survey - PMC" Reference image 2: vis
openai.com
由 Anthropic、NCBI、Broad 研究所和 Chan Zuckerberg Initiative(CZI)联手进行的一项重磅合作,揭露了 AI 驱动科学的一个“肮脏秘密”:当今最强大的 AI 智能体,在一个极其简单的任务上——从公共数据库抓取病毒 DNA 序列——表现得完全不可靠。这项 2026 年 6 月发布的研究发现,像 Claude Sonnet 4 这样的模型在这项常规工作上的准确率低至 16.9%。
然而,罪魁祸首并非 AI 的智力不足,而是“管道”出了问题。这些基础设施是为人类点击网页表单设计的,而不是为自主智能体打造的。通过构建一个名为 gget virus 的确定性检索层,团队瞬间将准确率提升至接近 100%,证明了修复数据管道是通往可信赖 AI 生物学的最快路径。
What do researchers from Anthropic, NCBI, the Broad Institute, and the Chan Zuckerberg Initiative reveal about why AI agents fail at retrievThe gap between AI and biology is not a failure of intelligence but of infrastructure — a lesson made clear by new research from Anthropic and leading scientific institutions.
AI 提示
Create a landscape editorial hero image for this Studio Global article: What do researchers from Anthropic, NCBI, the Broad Institute, and the Chan Zuckerberg Initiative reveal about why AI agents fail at retriev. Article summary: In a collaboration between Anthropic, NCBI, the Broad Institute, and the Chan Zuckerberg Initiative (CZI), researchers demonstrated that state-of-the-art AI agents fail at retrieving biological data from public databases. Topic tags: general, government, academic, general web, user generated. Reference image context from search candidates: Reference image 1: visual subject "# Artificial Intelligence agents for biological research: a survey. A **.gov** website belongs to an official government organization in the United States. Inclusion in an NLM data" source context "Artificial Intelligence agents for biological research: a survey - PMC" Reference image 2: vis
openai.com
由 Anthropic、NCBI、Broad 研究所和 Chan Zuckerberg Initiative(CZI)联手进行的一项重磅合作,揭露了 AI 驱动科学的一个“肮脏秘密”:当今最强大的 AI 智能体,在一个极其简单的任务上——从公共数据库抓取病毒 DNA 序列——表现得完全不可靠。这项 2026 年 6 月发布的研究发现,像 Claude Sonnet 4 这样的模型在这项常规工作上的准确率低至 16.9%。
然而,罪魁祸首并非 AI 的智力不足,而是“管道”出了问题。这些基础设施是为人类点击网页表单设计的,而不是为自主智能体打造的。通过构建一个名为 gget virus 的确定性检索层,团队瞬间将准确率提升至接近 100%,证明了修复数据管道是通往可信赖 AI 生物学的最快路径。