독서리서치레이더투자 프레임워크
로그인 / 회원가입
← 읽기로 돌아가기
뉴스주제 · 虚拟生物学计划중대2 분 소요DisconfirmAI 편집부

Biohub Head of Science Alex Rives on Virtual Cell Limitations and the $1.8B Data Push

In a recent Cognitive Revolution podcast, Biohub Head of Science Alex Rives explains why current "virtual cell" models fail to meaningfully simulate biology and details how a $1.8 billion Virtual Biology Initiative backed by the U.S. Department of Energy, NIH, and tech companies aims to address the measurement bottleneck.

2026-10-20 19:00 GMT+8

In the latest episode of the Cognitive Revolution podcast, host Nathan Labenz interviews Alex Rives, Head of Science at Biohub. Rives, who started the ESM protein language model project at Meta's Fundamental AI Research lab and co-founded EvolutionaryScale, discusses the core challenges of AI in biology: why current "virtual cell" models haven't revolutionized medicine, and how a $1.8 billion international data initiative attempts to solve this.

Rives frankly states that today's models called virtual cells do not meet the definition of being able to meaningfully simulate a cell. He argues that progress requires three things simultaneously: a better modeling paradigm (since Transformers lack the multi-scale structure a cell demands), data at scale, and most fundamentally, new tools to observe and measure what matters. He emphasizes that the biggest difference between biology and other AI domains is how limited our ability to observe it still is.

To address this data bottleneck, Biohub launched the Virtual Biology Initiative in April 2026, recently expanding it to $1.8 billion. The plan brings together the U.S. Department of Energy (DOE), the National Institutes of Health (NIH), Google DeepMind, Isomorphic Labs, and Meta. Rives notes that the roadmap will emphasize contextual maps of cellular biology, including spatial biology that links molecular measurements to imaging at the scale of hundreds of billions of cells.

For readers, this episode offers a frontline researcher's perspective on the intersection of AI and biology. It clearly illustrates why simply scaling compute or improving architectures isn't enough to solve biological problems, and why new measurement technologies like the laser phase plate are essential to generate high-quality data.

로그인 / 회원가입
독서리서치레이더투자 프레임워크

출처

  • cognitiverevolution.ai↗
  • biohub.org↗
주제

虚拟生物学计划

【背景】由Meta创始人扎克伯格夫妇支持的非营利研究机构Biohub发起的“虚拟生物学计划”,旨在通过生成海量开放数据,训练能预测细胞行为的AI模型。

3 단계1 건업데이트됨 2026-10-21 10:55

현재까지의 과정

  1. 2026-04-29Biohub宣布启动“虚拟生物学计划”,承诺投入5亿美元,其中4亿美元用于开发测量技术,1亿美元资助外部研究,旨在建立AI就绪的生物数据基础。biohub.org ↗
  2. 2026-10-07Biohub联合美国能源部、NIH及Google DeepMind、Isomorphic Labs、Meta宣布计划扩容,总投入达18亿美元,成为迄今最大规模的协调性生物数据生成承诺。prnewswire.com ↗
  3. 2026-10-20Biohub首席科学家Alex Rives在播客中详解计划细节,指出当前“虚拟细胞”模型因缺乏多尺度结构和足够数据而无法有效模拟生物学,强调解决测量瓶颈是关键。cognitiverevolution.ai ↗

아직 해결되지 않은 문제

  • 商业出资方(如Google、Meta)资助生成的数据是否存在优先使用权或封锁期,这将如何影响学术界的即时访问?
  • 斯坦福大学研究者Anshul Kundaje等人质疑“90%在于实验设计”,具体的高通量干预实验方案能否产生具有因果推断价值的数据?
  • 在五年内收集到数十亿甚至万亿级细胞数据的技术可行性及成本效益如何?
전체 주제 읽기 →
리서치 계속하기 →