John Jumper recounts winning the Nobel Prize for AlphaFold, from the anxious wait to the celebratory sparkling wine, and reflects on his unconventional path from dropping out of a physics PhD to leading one of biology's greatest breakthroughs.
要点 · TL;DR
AlphaFold 是有史以来最有用的 AI 工具，彻底改变了结构生物学和药物发现。 AlphaFold is the most useful AI tool ever, transforming structural biology and drug discovery.
AlphaFold 3 使用扩散架构处理蛋白质以外的多种生物分子。 AlphaFold 3 uses diffusion to handle diverse biomolecules beyond proteins.
AI 工具缩小了假设空间，使实验更高效。 AI tools narrow hypothesis space, making experiments more efficient.
核心观点 · Key points
AlphaFold 是有史以来最有用的 AI 工具，彻底改变了结构生物学和药物发现。 AlphaFold is the most useful AI tool ever, transforming structural biology and drug discovery.
AlphaFold 3 采用扩散架构，处理蛋白质以外的多种生物分子。 AlphaFold 3 uses a diffusion architecture to handle diverse biomolecules beyond proteins.
置信度指标对于可靠使用 AlphaFold 预测至关重要。 Confidence metrics are crucial for reliable use of AlphaFold predictions.
蛋白质设计进展迅速，但远未解决。 Protein design is advancing rapidly but still far from solved.
AI 工具缩小假设空间，使实验更高效。 AI tools narrow hypothesis space, making experiments more efficient.
反共识 · Contrarian takes
AlphaFold 的成功源于算力不足，而非充裕。 AlphaFold's success came from lack of compute, not abundance.
完美可解释性对于科学中有用的 AI 并非必需。 Perfect interpretability is not necessary for useful AI in science.
AlphaFold 2 的进化数据并不像假设的那样关键。 AlphaFold 2's evolutionary data was less critical than assumed.
解决重大挑战往往带来实用工具，而不仅仅是庆祝。 Solving grand challenges often leads to practical tools, not just celebration.
药物设计远不止结合；大多数药物在试验中失败。 Drug design is far more than just binding; most drugs fail in trials.
本期章节 · Chapters（共 15）
AlphaFold 的影响与诺奖AlphaFold's impact and Nobel Prize announcement
AlphaFold 的影响与意义AlphaFold's impact and significance
AlphaFold 的非常规用途Unusual uses of AlphaFold
AF2 与 AF3 的进化信息差异Evolutionary information in AlphaFold 2 vs AlphaFold 3
AF3 中的扩散与幻觉问题Diffusion in AlphaFold 3 and hallucination concerns
AI 与科学的可解释性Interpretability in AI and Science
AlphaFold 在疾病理解中的作用AlphaFold's role in understanding disease
从预测到设计：AlphaFold 的意外影响From prediction to design: AlphaFold's unexpected impact
AlphaProteio 与蛋白质设计AlphaProteio and Protein Design
工程生物学 vs. 预测Engineering Biology vs. Prediction
AI 与生物学：思考 vs. 实用AI and Biology: Thinking vs. Utility
迈向模拟细胞？Towards a Simulated Cell?
将 AlphaFold 与 LLM 集成Integrating AlphaFold with LLMs
生物学中难以计算的部分Aspects of biology resistant to computation