# MIRI The **Machine Intelligence Research Institute** is the Berkeley nonprofit that originated the modern AI existential-risk position. Founded in **2000** by Eliezer Yudkowsky as the **Singularity Institute for Artificial Intelligence**, backed initially by Brian and Sabine Atkins, it renamed to MIRI in January 2013. ## Why a small nonprofit gets a page MIRI has never been large, never shipped a product, and has published a modest research output. It is in this vault because of what it **manufactured**: the intellectual scaffolding, the community, and the vocabulary that a substantial part of the current AI industry now argues inside. The chain is traceable. Yudkowsky's writing on LessWrong built the **rationalist** community. That community supplied a large share of the early **[[Effective Altruism]]** movement's personnel and its framing of AI as the highest-leverage cause area. Effective altruism supplied money, staff and board members to frontier AI labs. The terminology the entire field now argues in — alignment, orthogonality, instrumental convergence, the treacherous turn — was largely produced or popularised in MIRI's orbit. There is an irony the vault should state plainly, because it is the most important thing about MIRI and it is not a criticism from outside: **the organisation founded to prevent the building of superintelligent AI is a significant part of the causal history of the labs racing to build it.** People radicalised into taking AGI seriously by MIRI's arguments concluded, in many cases, that the correct response was to build it first and carefully rather than not at all. ## The money, and why it matters here MIRI's funders map onto the vault's existing network files with unusual precision: **[[Peter Thiel]]** (early support, including the Singularity Summit from 2006), Open Philanthropy ($2.1M in 2019, $7.7M in 2020), and Vitalik Buterin (several million in Ethereum, 2021). Thiel money at the origin of the AI-risk movement, and Thiel money throughout the defense-tech and political networks documented elsewhere here, is the same money — which is a fact about the density of that network rather than evidence of a plan. ## The current posture MIRI has shifted decisively from technical alignment research toward **policy and public communication**. The 2025 book by Yudkowsky and Nate Soares — *If Anyone Builds It, Everyone Dies: Why Superhuman AI Would Kill Us All* — is the clearest statement of the position: not that alignment is hard, but that it will not be solved in time, and that the correct policy is to stop. That is a genuinely unusual place for an organisation to arrive: MIRI's public argument is now that its own founding research program failed, and that the remaining lever is political rather than technical. ## How the vault treats it MIRI belongs to the **[[AI Doomerism]]** tradition, which the vault holds as one of two live positions on AI harm — the other being the present-harms tradition (bias, labour, surveillance) that regards existential framing as itself a political move that draws scrutiny away from what is already happening. See the [[Artificial Intelligence]] gateway for that split. The vault takes neither as the neutral position. ## See Also - [[AI Doomerism]] · [[AI Safety]] · [[AGI]] — the intellectual tradition - [[Artificial Intelligence]] — the gateway, including the two-camps split - [[Effective Altruism]] — the movement MIRI's community seeded - [[Peter Thiel]] · [[Thiel Capital]] — early funding, and the wider network - [[Superintelligence (Book)]] — the argument's most-read popular statement