MIRI's Rob Bensinger and AI Risk Researcher EigenGender Clash Over ASI Doom Arguments
On September 14, MIRI researcher Rob Bensinger and @EigenGender, a researcher who works full-time on AI misalignment risk, held a multi-round exchange on X over ASI alignment. The core dispute is not whether AI is dangerous, but whether MIRI's arguments and confidence levels are justified. Neither side offered specific numerical estimates of doom probabilities, leaving the debate at the level of reasoning.
Confirmed
- Bensinger restated the classic Yudkowsky line: the most likely AI neither loves nor hates us, and indifference is fatal at capability limits—you are made of atoms it can use for something else.
- Bensinger framed the core disagreement: critics think modern ML is special and might accidentally yield a human-friendly ASI, whereas MIRI holds that modern ML is more dangerous in some ways and slightly better in others, without substantively reducing the difficulty of building aligned ASI.
- Bensinger added that the nuances of chapters 4-5 of If Anyone Builds It, Everyone Dies (IABIED) are hard to compress into tweets but do not change the bottom line: an ASI built this decade with modern ML is very unlikely to treat humanity well.
- EigenGender clarified that they work full-time on misalignment risk and are extremely worried about it; their objection concerns the argumentation, alleging Bensinger switches between distinct claims.
- EigenGender said IABIED never directly answers why such high confidence in doom is warranted, and found Bensinger's response unsatisfying.
- On the orthogonality thesis, EigenGender locates the fundamental disagreement in whether an arbitrary measure over mind-space exists: they believe it does not, and that on current paths we would likely stumble into an ASI that loves humanity; they also flag motte-and-bailey ambiguity in what the thesis means.
- EigenGender agrees modern ML will likely produce systems that don't love us, but holds the probability is not close to 1.
Unconfirmed
- Neither side provided numerical doom probabilities or their calculation basis; whether IABIED's arguments support the high confidence remains contested.
Why it matters
The debate reveals splits within the AI safety camp: even among researchers equally alarmed about misalignment, there are fundamental disagreements over the orthogonality thesis, whether modern ML changes alignment difficulty, and whether doom probability approaches 1—arguments that shape how the public and policymakers judge the credibility of AI doom claims.
2026-09-14 ~ 2026-09-14 · 7 related posts
Primary sources
- Alignment Debate: Indifferent AI Is Fatal, or Will ASI Love Us by Accident? — EigenGender · 2026-09-14
- [source] The orthogonality thesis debate: where alignment skeptics and MIRI actually diverge — EigenGender · 2026-09-14
- MIRI doubles down: ASI from modern ML this decade is very unlikely to love us — robbensinger · 2026-09-14
- MIRI's Rob Bensinger clashes over IABIED's p(doom) certainty — EigenGender · 2026-09-14
- [source] MIRI responds to critics: modern ML doesn't dent the difficulty of aligning ASI — robbensinger · 2026-09-14
- [source] Alignment researcher: I'm terrified of misalignment, but MIRI is conflating claims — EigenGender · 2026-09-14
- Full-time alignment researcher pushes back: misaligned AI likely, but not near-certain — EigenGender · 2026-09-14