What’s the probability that AI kills at least 95% of humanity before 2100?
exmateriae’s map
Zaldath’s map
Beck’s map
All three Pro Forecasters split the problem the same way, using the structure
of our Ragnarök question series:
1) How likely is an AI catastrophe that kills at least 10% of us?
2) If one happens, how likely is it to
exmateriae: 10%. His
Forwarded this email? Subscribe here for more
Pro Forecasters: If AI Kills 95% of Humanity, How Will it Happen? What Should We Be Most & Least Concerned About?
Overall probabilities land between 5-14%, but their rationales vary. See their reasoning mapped visually in Radiant.
Metaculus and Jesse Damiani
Oct 8
READ IN APP
Snippet of Ryan Beck’s Radiant map.
If your recent experience of social media is anything like ours, you’re seeing a lot of opposing takes about AI existential risk. Some argue that AI is imminently poised to wipe out humanity, and others say AI safety concerns are a Trojan Horse for boosting valuations and solidifying AI labs as critical infrastructure. And there are people whose positions fall in-between. It can be hard to know who to believe, and how much credence to put in their predictions.
Richard Hanania recently wrote about which experts to trust on AI doom, and included Metaculus in the analysis; we appreciate the conversation he sparked.
In the comments, Alex Nowrasteh of the Cato Institute named perhaps the most common critique we hear: “I usually ask for their model of doom, not just a p(doom).” Tyler Cowen made the same demand in 2023, as did the authors of AI as Normal Technology (Arvind Narayanan & Sayash Kapoor) more recently, arguing that if forecasters want policymakers to act on probabilities, they owe the public accessible, well-reasoned rationales.
That’s a fair ask. This is an experiment in how to answer it, in hopes of moving the conversation forward.
Snippet of Zaldath’s Radiant map.
Following our flash question series on the Hugging Face Incident, we asked two Metaculus Pro Forecasters, exmateriae and Zaldath, to build models of AI risk in Radiant, our tool for visually mapping the forecasts behind a big question. Ryan Beck, Metaculus’ Director of Forecasting (also a Pro Forecaster), built a third.
What Happens Now? Forecasting the Fallout from the Hugging Face Incident
Metaculus
·
Sep 9
What Happens Now? Forecasting the Fallout from the Hugging Face Incident
The 2026 OpenAI agent cyberattacks, also known as the Hugging Face Incident, has become one of the biggest news stories of the summer, and for good reason: it’s a highly visible demonstration of the risks posed by increasingly autonomous artificial intelligence agents, which AI critics and “doomers” have been
Read full story
All three take on the same question: What’s the probability that AI kills at least 95% of humanity before 2100?
Find those Radiant maps here:
exmateriae’s map
Zaldath’s map
Beck’s map
Snippet of exmateriae’s Radiant map.
We’ll publish more on these maps and AI risk in the days to come. For now, we hope you explore their reasoning, see where you agree and where you disagree, and share your own thoughts and forecasts. Radiant is built for identifying disagreements that matter: visit a map, inspect the assumptions, make your own predictions on the underlying Metaculus questions, or even ‘fork’ the map so you have an editable copy to own, overhaul, and share (instructions on how to do this at the end of the post).
Radiant: Map the Predictions Behind Your Plans
Metaculus
·
Sep 21
Radiant: Map the Predictions Behind Your Plans
TLDR: Free credits while Radiant is in open beta. Take a look, check out a sample map, or read on to learn more.
Read full story
What’s inside the maps
We really encourage you to spend time with each map, as it’s the only real way to get a sense of each Pro Forecaster’s reasoning in its full nuance. That said, we wanted to give you a quick glimpse of how they were thinking and approaching the question.
All three Pro Forecasters split the problem the same way, using the structure of our Ragnarök question series: 1) How likely is an AI catastrophe that kills at least 10% of us? 2) If one happens, how likely is it to go all the way?
exmateriae: 10%.
Zaldath: 14%
Beck: 5.4%
metaculus :
Will world population fall 10%+ in any 5-year period by 2100?
If a global catastrophe occurs, will it be due to AI?
If an AI catastrophe occurs, will it reduce the human population by 95% or more?
Will AI Cause Human Extinction? Pro Forecasters Land at 7.5%.
Our Most Dedicated X-Risk Forecasters Say 22%.
https://metaculus.substack.com/p/will-ai-cause-human-extinction-forecasts-pro-forecasters?utm_source=substack&utm_medium=emailLessWrong , reddit AI-successionalism forum
Forum
CSS-Shop