AI Safety must industrialize: Vannevar Bush's Manhattan Project lessons apply
Urgent push for active senior researcher recruitment, not just grant applications.
In a LessWrong essay crossposted from Canary Institute, dan.parshall draws parallels between the early Manhattan Project's institutional apathy and today's AI Safety field-building challenges. He recounts how British reports on uranium requirements were ignored by the American committee chair, until physicist Mark Oliphant personally traveled to the US to force action. The true breakthrough came when Vannevar Bush secured presidential authority and recruited senior academics to mobilize the community—a model parshall believes AI Safety needs today, urgently transitioning from a "community" to an "industry."
Parshall critiques the current pull-based system, where researchers wait for grant applications (like RFPs), instead of pushing to recruit established experts. He notes that while the AI-verification literature has adopted the "covert adversary" framework from Aumann and Lindell, no funder seems to have approached these authors to pivot their work. Similarly, he highlights Palisade's success in making AI timelines legible to policymakers as a rare example of effective proactive engagement. With a major funding wave heading toward AI Safety, pull-model machinery will fail to absorb it; the field needs senior academic groups incentivized to change focus, not just individual residents in small programs.
- Manhattan Project succeeded only after Vannevar Bush personally recruited senior academics, not via passive grant applications.
- AI Safety funders have not proactively approached pioneers like Aumann and Lindell despite adopting their covert-adversary framework.
- Palisade's policy engagement is a good model; similar effort is needed for academic researchers before a large funding wave hits.
Why It Matters
Proactive recruitment of senior researchers is critical for AI Safety to scale and absorb upcoming funding.