Podcast Atlas

AI4S 需要狂人与野心家｜对话英灵殿 Odin："如果神存在，我怎能容忍自己不是神？"【公路播客】

source Episode summary Updated 2026-07-24 Tags: Podcast, Ai-for-Science, Biotech, Drug-Discovery

Summary

This Shizilukou Crossing road-podcast episode interviews [[HaotianOdin|Odin / 浩天]], founder of [[YinglingdianAI|英灵殿]], about building an AI For Science company whose near-term product is an AI Drug Discovery Platform and whose long-term ambition is Scientific Discovery Automation. The discussion centers on an All-Modal Molecular World Model that would model and design small molecules, proteins, RNA, and DNA together rather than treating each molecular modality as a separate field. The second half turns the same technical ambition into a founder-method interview: Haotian Odin / 浩天 explains his move from Baker Lab, his preference for platform over drug pipelines, and his attempt to preserve Founder Signal Discipline under financing pressure.

Key Claims

  • Yinglingdian AI / 英灵殿 is presented as a roughly 30-person company with tens of millions of dollars in financing, a short-term goal of becoming a strong AI Drug Discovery Platform, and a longer goal of moving toward general scientific AI.
  • Haotian Odin / 浩天 frames science acceleration as automating data analysis, hypothesis generation, and path selection, making Scientific Discovery Automation the episode’s highest-level technical thesis.
  • The company’s core technical bet is an All-Modal Molecular World Model: one framework that can reason across DNA, RNA, proteins, and small molecules because life processes depend on interactions among all of them.
  • [[ODesign|O-Design / Odyssey]] is described as an early open-source exploration where one model could attempt small-molecule, RNA, DNA, and protein design, with simple downstream validation and undergraduate-level participation in multiple targets.
  • [[LDesign|L-Design]] is introduced as a coming milestone closer to a molecular world model, with examples such as designing proteins from small molecules, small molecules from small molecules, and DNA aptamers from proteins.
  • AlphaFold and diffusion-style protein design appear as enabling precedents, but Haotian Odin / 浩天 argues that architecture should serve the scientific problem rather than become an object of attachment.
  • Baker Lab matters as both technical formation and exit contrast: the source says Haotian Odin / 浩天 learned from AI protein design there but left because he wanted to pursue his own technical path rather than become only an engineer inside a famous lab.
  • The source treats Platform-Pipeline Biotech Strategy as a live business decision. Yinglingdian AI / 英灵殿 does not plan to build drug pipelines in the short term because that could distract the AI team and turn pharmaceutical customers into competitors, while still acknowledging that pipelines can prove platform value.
  • Domain Expert Alignment remains necessary. The source emphasizes AI-native researchers, CS/model-training backgrounds, experimental collaboration, commercial interfaces, and willingness to learn over narrow labels.
  • Life Reprogramming Vision is the episode’s outer horizon: intervening in interactions among DNA, RNA, proteins, and small molecules is framed as reprogramming life, while the source also treats life creation and consciousness questions as distant and ethically charged.
  • Personalized Molecular Medicine is presented as the more reachable ordinary-person milestone: custom drugs based on genotype and protein-expression state rather than one-size-fits-all treatment.
  • Founder Signal Discipline is the founder-method layer. Haotian Odin / 浩天 describes entrepreneurship as arising from dissatisfaction with the world, then uses Buddhist practice language, “be yourself,” and resistance to VC-friendly buzzword stitching to explain how he tries to keep the company aligned with its original scientific question.

Key Quotes

“科学发现的蒸汽机” — Haotian Odin / 浩天’s metaphor for AI for Science.

“如果神存在的话，那么我又如何能够容忍自己不成为一个神” — the source’s title quote, linked to molecular creation ambition.

“这是好事” — the founder’s shorthand for rejecting a fixed good/bad frame.

“何为祖师西来意” — the question he says he would ask Mark Zuckerberg as a test of whether original intention survives success.

Connections

Contradictions

  • No direct contradiction with prior wiki content.
  • The source extends existing AI For Science and Discovery Model material by adding a biological-molecule route where experimental feedback, molecular modality unification, and platform-vs-pipeline choices are as important as model capability.
  • The source qualifies broad World Models enthusiasm by grounding “world model” in molecular interaction and experimental validation rather than visual simulation or generic next-state prediction.