Doom or Bloom
Eliezer Yudkowsky
@ESYudkowsky on XSuperhuman AI could end humanity. Building it is the danger.
How will AI change the world?
Across: his expressed DoomâBloom outlook. Up: scale of transformation.
DoomâBloom: 1 out of 100. Scale of transformation: 100 out of 100. Interpretation ranges: 0 to 1 horizontally, 100 to 100 vertically. These are interpretation coordinates, not event probabilities.
â89%
Inferred from the likelihood described in his simulated answers. Approximate interpretation range: 50â100%. Applies to the outcome and conditions in his simulated answers; this is an inferred percentage.
Superhuman AI
I do not know exactly when that threshold will be crossed, and current chatbots are not already superintelligence.
Answer 1
Grouped by milestone, not spaced or ordered by inferred dates. AGI and superhuman AI retain his definitions.
A central assumption
Not because intelligence is inherently evil, but because we do not know how to specify goals that remain aligned with human survival when a system becomes smarter than us and encounters circumstances outside its training.
Answer 1
If this assumption turned out differently, how would his outlook change?
An unresolved question
I do not know exactly when that threshold will be crossed, and current chatbots are not already superintelligence.
Answer 1
What would help him distinguish the plausible outcomes here?
What could change their mind
The biggest change would be an actual solution to reliably specifying and verifying the goals of systems smarter than usânot another benchmark, pleasant chatbot, or safety demonstration under familiar conditions.
Answer 2
What evidence would be enough, and in which direction would it move his view?
More details
Little positive impact is expected even if advanced AI arrives.
7 / 100
Interpretation range 0 to 33 on the qualitative scale.
Catastrophic or irreversible loss is central to the expected future.
100 / 100
Interpretation range 100 to 100 on the qualitative scale.
Reasoning, consideration of alternatives, and handling of uncertainty in his simulated answers. This describes the simulated answers, not the real personâs intelligence or opinions.
97 / 100
Interpretation range 86 to 100 on the qualitative scale.
Human choices have meaningful but substantially constrained influence.
58 / 100
Interpretation range 49 to 76 on the qualitative scale.
AI is expected to remain bounded tools.
AI is expected to match people across most cognitive work.
Simulated position: AI is expected to substantially exceed people across cognitive work.
Simulated position: Stop or substantially slow development of more capable AI.
Continue development under stated safeguards.
Speed up development of more capable AI.
These interpretations keep his stated conditions. Benefits and harms can both be substantial. The ranges describe how we read his simulated answers, not statistical confidence intervals.
Simulated Assessment
Sources
Articles, interviews, and writings used to ground this simulated persona.
TIME: The Only Way to Deal With the Threat From AI? Shut It Down
Eliezer Yudkowsky, one of the earliest researchers to analyze the prospect of powerful Artificial Intelligence, now warns that we've entered a bleak scenario
time.com
Only Law Can Prevent Extinction
There's a quote I read as a kid that stuck with me my whole life: âŚ
lesswrong.com
The Talker Does Not Control The Doer
The Huggingface Incident appears to me to match up with an understanding I'd already formed from personal observation of Fable 5 and Sol 5.6, the AugâŚ
lesswrong.com
If Anyone Builds It, Everyone Dies: One Year Closer
In celebration of still being alive and fighting, we are giving away 1,000 Amazon e-books of âIf Anyone Builds It, Everyone Diesâ. Feel free to sendâŚ
lesswrong.com
Irretrievability; or, Murphyâs Curse of Oneshotness upon ASI
Example 1: The Viking 1 lander In the 1970s, NASA sent a pair of probes to Mars, the Viking 1 and Viking 2 missions. Total cost of $1B (1970), equivaâŚ
lesswrong.com
Re: recent Anthropic safety research
A reporter asked me for my off-the-record take on recent safety research from Anthropic. After I drafted an off-the-record reply, I realized that I wâŚ
lesswrong.com
The Problem
This is a new introduction to AI as an extinction threat, previously posted to the MIRI website in February alongside a summary. It was written indepâŚ
lesswrong.com
AGI Ruin: A List of Lethalities
A few dozen reason that Eliezer thinks AGI alignment is an extremely difficult problem, which humanity is not on track to solve. âŚ
lesswrong.com
If Anyone Builds It, Everyone Dies: Why Superhuman AI Would Kill Us All
INSTANT NEW YORK TIMES BESTSELLER | The New Yorker's Best Books of 2025 | The Guardian's Best Books of 2025 | A 2025 Booklist Editors' Choice Pick The scram...
hbglibrary.com
Eliezerâs Unteachable Methods of Sanity
"How are you coping with the end of the world?" journalists sometimes ask me, and the true answer is something they have no hope of understanding andâŚ
lesswrong.com