AlphaFold 获得诺贝尔奖两年后，其核心研发团队面临大规模人事调整。[1] AlphaFold 论文的大部分原始作者在过去一年被重新分配，全职核心作者中近四分之一已离开 Google。[1] 这一现象反映了一个更广泛的趋势：全球顶尖教授、数学家和物理学家正在大规模流向 Anthropic、OpenAI、DeepMind 等 AI 企业。
人才转向企业的吸引力主要来自薪酬差异。[1] 企业中顶尖 1% 的 AI 研究者年收入从 59.5 万美元上升至 194 万美元，而大学同梯队研究者的收入仅从 30.1 万美元增至 39.2 万美元。[1] 数据显示，2022 年 AI 博士毕业后进入企业的比例已升至 70.7%，进入学术界的比例仅为 20%。[1]
这种人才流失带来了深刻的科研格局变化。[1] 2010 年代初约 65% 的大规模机器学习模型由学术实验室开发，到 2022 年这一比例逆转，约 81% 的前沿模型由企业独立完成。[1] 从 2000 年至 2019 年，企业 AI 研究者占比从 48% 升至 68%，而论文占比仅从 27% 升至 32%，但专利占比已从 86% 升至 95%。[1] NBER 研究发现，AI 学者长期转入企业后，论文产量平均减少 65%。[1] 此外，2024 年新毕业生仅占大型科技公司招聘人数的 7%，相比 2019 年下降超过一半。[1]
Two years after AlphaFold's Nobel Prize recognition, the foundational shift in scientific research has become undeniable: elite researchers are leaving academia en masse for artificial intelligence companies.[1] The core team behind AlphaFold has been dismantled and reorganized, with nearly a quarter of its full-time core authors departing Google.[1] This exodus reflects a broader trend in which top-tier professors, mathematicians, and physicists are concentrating at firms like Anthropic, OpenAI, and DeepMind, fundamentally reshaping the landscape of scientific discovery.
The scale of talent migration is striking. Among AI PhD graduates, the proportion entering industry climbed from 40.9% in 2011 to 70.7% in 2022, while those entering academia fell to just 20%.[1] Compensation disparities amplify the pull: top 1% AI researchers at enterprises saw annual earnings surge from $595,000 to $1.94 million, whereas their university counterparts experienced only a modest rise from $301,000 to $392,000.[1] As a consequence, enterprises have consolidated control over cutting-edge model development, computational resources, and research agendas. The share of frontier AI models developed independently by companies reached approximately 81% by 2022, up from roughly 65% in the early 2010s when academic laboratories dominated.[1]
Universities face a mounting crisis. Research by the National Bureau of Economic Research found that when AI scholars transition to industry, their paper output declines by an average of 65%.[1] The sectoral shift is reflected in hiring patterns: newly graduated PhD holders comprised only 7% of large technology companies' recruitment in 2024, a decline of more than half from 2019.[1] Between 2000 and 2019, while the share of AI researchers in enterprise grew from 48% to 68%, their paper output rose only from 27% to 32%, even as their patent share surged from 86% to 95%.[1] The disparity underscores a troubling reality: as commercial incentives drive research toward patentable and productizable advances, fundamental scientific inquiry withers in institutional settings starved of talent and resources.
