AI enthusiasts love to say that the technology is as revolutionary and important as nuclear weapons. Even the Trump administration has adopted the metaphor. The President and the Department of Energy have repeatedly referred to the development of AI in the US as “Manhattan Project 2.0.”
But is the buildout of LLMs and machine learning systems really as important as the development of the atom bomb? And what are the lessons from the atomic age that AI scientists should then learn? Do we need an AI Non Proliferation Treaty? An AI International Atomic Energy Agency?
On this episode of Angry Planet, Ankit Panda comes on to talk about the uses and limitations of the “AI as nuclear weapons” metaphor. Panda is an expert in nukes and a Senior Fellow at the Carnegie Endowment for International Peace. He’s been sharing his extended thoughts on the AI-nuclear connection at his Nukesletter Substack.
Stanislav Petrov
AI as nuclear weapons
Why nuclear weapons resonate with people in the AI field
The Strategic Air Command story
That time we spilled nuclear material all over Greenland and Spain
NNSA and Anthropic
AI as the next Manhattan Project
A massive infrastructure project
Fissile material as silicon
What’s the AI version of an NPT and IAEA?
AI and nuclear are both dual use
On AI winters
What AI is actually being used for, what it might be used for
The socialization around AI will change.
AI Arms and Influence: Frontier Models Exhibit Sophisticated Reasoning in Simulated Nuclear Crisis
