Back
AGI Ruin: A List of Lethalities
blogAuthor
Credibility Rating
3/5
Good(3)Good quality. Reputable source with community review or editorial standards, but less rigorous than peer-reviewed venues.
Rating inherited from publication venue: Alignment Forum
A widely-cited and debated 2022 post by Eliezer Yudkowsky representing the strongest public statement of his doom thesis; essential reading for understanding the pessimistic wing of AI safety discourse and the arguments that motivate MIRI's research priorities.
Metadata
Importance: 90/100blog postprimary source
Summary
Eliezer Yudkowsky's comprehensive argument for why AGI development is likely to result in human extinction, presented as a list of distinct failure modes and reasons why alignment is extremely difficult. The post systematically addresses why standard proposed solutions are insufficient and why the default outcome of unaligned AGI is catastrophic. It serves as a canonical statement of Yudkowsky's pessimistic position on humanity's ability to navigate the AGI transition safely.
Key Points
- •Lists dozens of independent 'lethalities'—reasons why AGI development leads to doom even if individual problems seem solvable, emphasizing cumulative difficulty.
- •Argues that outer alignment, inner alignment, and interpretability are each individually insufficient and collectively still likely to fail under real AGI development conditions.
- •Contends that current ML paradigms produce systems whose internals are opaque, making verification of alignment nearly impossible before deployment at dangerous capability levels.
- •Challenges optimistic views that iterative deployment, scaling feedback, or governance can compensate for fundamental alignment uncertainty.
- •Represents Yudkowsky's explicit claim that without major breakthroughs in alignment theory, AGI timelines imply near-certain catastrophe regardless of developer intent.
Cited by 2 pages
| Page | Type | Quality |
|---|---|---|
| Sharp Left Turn | Risk | 69.0 |
| AI Doomer Worldview | Concept | 38.0 |
Cached Content Preview
HTTP 200Fetched Apr 7, 202698 KB
Oct
NOV
Dec
12
2024
2025
2026
success
fail
About this capture
COLLECTED BY
Collection: Common Crawl
Web crawl data from Common Crawl.
TIMESTAMPS
The Wayback Machine - http://web.archive.org/web/20251112111245/https://www.alignmentforum.org/posts/uMQ3cqWDPHhjtiesc/agi-ruin-a-list-of-lethalities
This website requires javascript to properly function. Consider activating javascript to get access to all site functionality.
1964
AI ALIGNMENT FORUM
AF
Login
1963
AGI Ruin: A List of Lethalities — AI Alignment Forum
2022 MIRI Alignment Discussion
AI RiskThreat Models (AI)Double-CruxFuzziesLanguage Models (LLMs)Meetups & Local Communities (topic)AI
Curated
147
spotify-podcast-badge-wht-blk-165x40Created with Sketch.
AGI Ruin: A List of Lethalities
by Eliezer Yudkowsky
5th Jun 2022
36 min read
711
147
AI RiskThreat Models (AI)Double-CruxFuzziesLanguage Models (LLMs)Meetups & Local Communities (topic)AI
Curated
Previous:
Six Dimensions of Operational Adequacy in AGI Projects
33 comments317 karma
Next:
A central AI alignment problem: capabilities generalization, and the sharp left turn
18 comments276 karma
Log in to save where you left off
AGI Ruin: A List of Lethalities
77evhub
13Eliezer Yudkowsky
74evhub
34Eliezer Yudkowsky
12Alexander Gietelink Oldenziel
10Vaniver
2Eliezer Yudkowsky
5James Payor
7James Payor
36Matthew Barnett
13Vaniver
5Buck
12Daniel Kokotajlo
7leogao
29Rob Bensinger
0The Dao of Bayes
7Rob Bensinger
27Vanessa Kosoy
4Rob Bensinger
9Vanessa Kosoy
4Rob Bensinger
3Vanessa Kosoy
1Rob Bensinger
1Vanessa Kosoy
25Wei Dai
25Steven Byrnes
12CarlShulman
24John Schulman
11Vaniver
3John Schulman
3Vaniver
9Eliezer Yudkowsky
8John Schulman
15Vaniver
23TurnTrout
22Daniel Kokotajlo
5Rob Bensinger
-12Eliezer Yudkowsky
5Rafael Harth
3Rob Bensinger
22johnswentworth
2Ruby
15johnswentworth
14evhub
4Eliezer Yudkowsky
4evhub
12Steven Byrnes
0The Dao of Bayes
0Rob Bensinger
0The Dao of Bayes
3Rob Bensinger
21ESRogs
11Steven Byrnes
5Rob Bensinger
3James Payor
3James Payor
20romeostevensit
3Ben Pace
16Raemon
15Andrew_Critch
11Vika
11habryka
3Eliezer Yudkowsky
10Richard_Ngo
6Eliezer Yudkowsky
4Richard_Ngo
3Eliezer Yudkowsky
3Richard_Ngo
8Eliezer Yudkowsky
9p.b.
0Rob Bensinger
9p.b.
6Eliezer Yudkowsky
6TurnTrout
3Rob Bensinger
6TurnTrout
6Rob Bensinger
4TurnTrout
2Vaniver
2TurnTrout
2lc
2TurnTrout
1lc
2TurnTrout
1lc
3TurnTrout
2FireStormOOO
2TurnTrout
8William_S
5Rob Bensinger
7Raemon
6Evan R. Murphy
6David Scott Krueger (formerly: capybaralet)
2Rob Bensinger
1David Scott Krueger (formerly: capybaralet)
6Vaniver
6DaemonicSigil
8Ramana Kumar
9Eliezer Yudkowsky
31DaemonicSigil
19Eliezer Yudkowsky
2DaemonicSigil
1Edouard Harris
6AlphaAndOmega
1sullyj3
19Rob Bensinger
... (truncated, 98 KB total)Resource ID:
0aea2d39b8284ab1 | Stable ID: sid_WAssQPQdLi