跳到正文
Google DeepMind· @GoogleDeepMind · X·· 6 天前AI 评分58
AI 导读

Google DeepMind 发文介绍一款模型的三项能力:驱动 AlphaProof Code 智能体在数周内完成 Signal 加密代码的数学安全证明,而非数年;直接从实验记录本和原始数据中发现湿实验的细微错误与数据异常;在大型数据库中找出多出 40 倍的相关蛋白质序列,用于辅助建模难以预测的蛋白质。

正文 · 原文

🔘 It powered AlphaProof Code agents to mathematically prove the security of @Signalapp’s cryptographic code in weeks rather than years.

🔘 Spotted subtle wet-lab errors and data anomalies directly from lab notebooks and raw data.

🔘 Found 40 times more related protein sequences across large databases to help model hard-to-predict proteins.

来源:Google DeepMind · x.com