How AI Experts Forecast Doom Probabilities

Introduction

Surveys of AI researchers and related experts are among the most tangible evidence we have about how people deeply involved in developing and studying artificial intelligence forecast the probability of existential catastrophe—the chance that future AI systems might cause human extinction, permanent loss of human autonomy, or similarly irreversible harm. Rather than being single authoritative predictions, these surveys reveal a broad distribution of subjective probability estimates (often called “p(doom)”), shaped by how questions are framed, who is asked, and how respondents interpret unfamiliar future scenarios. [AI Wiki]wiki.aiimpacts.orgAI Impacts Wiki Surveys of experts on levels of AI Risk [AI Impacts WikiAI Impacts WikiSurveys of experts on levels of AI Risk [AI Impacts Wiki]May 9, 2023…Published: May 9, 2023

Expert surveys matter because they give a structured snapshot of current beliefs across technical communities. They help policymakers, journalists, and the public see where experts disagree, why the disagreements arise, and what assumptions underlie different probability forecasts. They also highlight the limits of expert intuition when applied to unprecedented technological change. [IEEE Spectrum]spectrum.ieee.orgSpectrum Weighing the Prophecies of AI DoomIEEE SpectrumWeighing the Prophecies of AI Doom - IEEE SpectrumJanuary 25, 2024…Published: January 25, 2024

Survey Methodologies and Respondent Selection

A key complexity in interpreting AI doom surveys is who is being asked and how.

Large communitywide polls:

The best‑known recent survey was conducted by researchers at AI Impacts in 2023. It sampled 2,778 researchers who had published at leading AI conferences and journals, not just specialists in safety or existential risk. Participants were asked about timelines for transformative AI and the likelihood of “human extinction or similarly permanent and severe disempowerment” from future AI advances within 100 years. [AI Wiki]wiki.aiimpacts.orgAI Impacts Wiki Surveys of experts on levels of AI Risk [AI Impacts WikiAI Impacts WikiSurveys of experts on levels of AI Risk [AI Impacts Wiki]May 9, 2023…Published: May 9, 2023

Smaller expert subsets:

Separate, smaller surveys and studies focus specifically on AI safety researchers or experts with self‑described familiarity with existential risk concepts. These groups tend to include people working at safety institutes, alignment research labs, or philosophy departments concerned with long‑term risk. For example, an analysis published in 2025 surveyed 111 such experts to understand the roots of disagreement about p(doom), including technical assumptions like takeoff speed and the difficulty of alignment. [arXiv]arxiv.orgarXivWhy do Experts Disagree on Existential Risk and P(doom)? A Survey of AI ExpertsJanuary 25, 2025…Published: January 25, 2025

[Hybrid forecasting tournaments:]sciencedirect.comScienceDirectSubjective-probability forecasts of existential risk: Initial results from a hybrid persuasion-forecasting tournament - Scie…

Some work has experimented with structured forecasting methods, bringing in “superforecasters” and specialist groups to combine elicitation with peer discussion. These studies explore not just what probabilities experts give but how confident they are, where they anchor their judgments, and how those forecasts evolve in dialogue. [ScienceDirect]sciencedirect.comScienceDirectSubjective-probability forecasts of existential risk: Initial results from a hybrid persuasion-forecasting tournament - Scie…

These differing survey populations and methods substantially influence outcomes: broader AI research communities include many people unfamiliar with alignment concepts, while specialist safety communities tend to yield higher median risk forecasts. AI Impacts Wiki

Differences in Question Framing and Definitions

A persistent challenge when interpreting p(doom) surveys is that “doom” and the question context vary from one instrument to another:

• Some questions explicitly ask about human extinction, while others include broader outcomes such as “permanent and severe disempowerment of the human species.” [AI Wiki]wiki.aiimpacts.orgAI Impacts Wiki Surveys of experts on levels of AI Risk [AI Impacts WikiAI Impacts WikiSurveys of experts on levels of AI Risk [AI Impacts Wiki]May 9, 2023…Published: May 9, 2023
• Time horizons differ: many surveys ask about risks over the next century, others leave the timeframe implicit. Differences in horizons can change how respondents anchor their subjective probabilities. [AI Wiki]wiki.aiimpacts.orgAI Impacts Wiki Surveys of experts on levels of AI Risk [AI Impacts WikiAI Impacts WikiSurveys of experts on levels of AI Risk [AI Impacts Wiki]May 9, 2023…Published: May 9, 2023
• Definitions of the underlying AI capability also vary: some surveys present the hypothetical of machines that can outperform humans in every task, while others leave capabilities unspecified. Critics have pointed out that framing assumptions can bias responses toward higher perceived risk if respondents take transformative AI as a given. [Scientific American]scientificamerican.comScientific AmericanAI Survey Exaggerates Apocalyptic Risks | Scientific AmericanJanuary 26, 2024…Published: January 26, 2024

This variation in framing is not just academic; it directly affects the numeric estimates experts provide. In the AI Impacts survey, respondents were first asked to assume a scenario of “very advanced AI” when estimating existential risk, a design choice that some critics believe implicitly nudges respondents toward non‑zero risk estimates. [Scientific American]scientificamerican.comScientific AmericanAI Survey Exaggerates Apocalyptic Risks | Scientific AmericanJanuary 26, 2024…Published: January 26, 2024

Distribution of p(doom) Estimates Across Experts

Despite methodological differences, several consistent patterns emerge from expert surveys:

1. Median and mean estimates among broad AI researchers:

The 2023 AI Impacts survey’s most widely cited results show a median probability of ~5% and a mean of ~14.4% for AI leading to extinction or similarly severe outcomes within 100 years. This reflects a long tail of higher estimates from a minority of respondents. [AI Wiki]wiki.aiimpacts.orgAI Impacts Wiki Surveys of experts on levels of AI Risk [AI Impacts WikiAI Impacts WikiSurveys of experts on levels of AI Risk [AI Impacts Wiki]May 9, 2023…Published: May 9, 2023

2. Specialist safety and governance researchers tend to predict higher risk:

Smaller surveys that concentrate on people working directly on AI safety and long‑term risk often report median p(doom) estimates in the 20–30% range and even higher in some measures of concern that emphasise lack of technical alignment research or governance. [Longterm Wiki]longtermwiki.comLongterm Wiki AI Accident Risk Cruxes | Longterm WikiLongterm WikiAI Accident Risk Cruxes | Longterm WikiFebruary 28, 2026…Published: February 28, 2026

3. Wide individual variance:

Even within surveys, responses vary dramatically. A summary of expert estimates—drawing on multiple sources—ranges from near zero to probabilities above 90% for existential catastrophe, depending on individual views of future AI capabilities and control challenges. [AI Safety Atlas]ai-safety-atlas.comAI Safety Atlas Appendix: Quantifying Existential RisksAI Safety AtlasAppendix: Quantifying Existential Risks - Chapter 2 - AI Safety Atlas…

4. Disagreement drivers identified by researchers:

Studies that ask experts why they disagree find that differences often come down to assumptions about AI takeoff speed, the feasibility of alignment solutions, and how society will govern powerful AI systems. Technical background alone does not predict higher or lower doom probabilities as strongly as these deeper beliefs about future trajectories. [AI Models]aimodels.fyiAI ModelsWhy do Experts Disagree on Existential Risk and P(doom)? A Survey of AI Experts | AI Research Paper DetailsFebruary 24, 2025…Published: February 24, 2025

What Expert Surveys Reveal — and What They Don’t

Taken together, expert surveys do not produce a single consensus number for p(doom). Instead they show:

• A non‑trivial subset of researchers assign non‑zero probability to existential catastrophe, even when controlling for methodological variation. [AI Wiki]wiki.aiimpacts.orgAI Impacts Wiki Surveys of experts on levels of AI Risk [AI Impacts WikiAI Impacts WikiSurveys of experts on levels of AI Risk [AI Impacts Wiki]May 9, 2023…Published: May 9, 2023
• Estimate distributions are long‑tailed and sensitive to framing, definitions, and respondent self‑selection. [Scientific American]scientificamerican.comScientific AmericanAI Survey Exaggerates Apocalyptic Risks | Scientific AmericanJanuary 26, 2024…Published: January 26, 2024
• The surveys reflect not just beliefs about technical likelihoods, but interpretive lenses about AI futures, governance effectiveness, and the nature of intelligence itself. [AI Models]aimodels.fyiAI ModelsWhy do Experts Disagree on Existential Risk and P(doom)? A Survey of AI Experts | AI Research Paper DetailsFebruary 24, 2025…Published: February 24, 2025

Importantly, expert surveys are not predictive models. They are snapshots of current beliefs under deep uncertainty, useful for understanding what experts think and disagree about, not for deriving precise probabilities of future outcomes.

Implications of Survey Evidence

For readers trying to understand the significance of expert surveys in the context of AI doom debates:

• Variation is normal: Disagreement among experts is expected when forecasting unprecedented technological and social change. [AI Models]aimodels.fyiAI ModelsWhy do Experts Disagree on Existential Risk and P(doom)? A Survey of AI Experts | AI Research Paper DetailsFebruary 24, 2025…Published: February 24, 2025
• Population matters: Views differ between general AI researchers and those focused on safety or governance, suggesting that estimates incorporate attitudinal as much as technical differences. [Longterm Wiki]longtermwiki.comLongterm Wiki AI Accident Risk Cruxes | Longterm WikiLongterm WikiAI Accident Risk Cruxes | Longterm WikiFebruary 28, 2026…Published: February 28, 2026
• Framing shapes responses: Careful attention to how questions are asked reveals that numeric p(doom) estimates are about more than just risk, they are about interpretation and communication of uncertain futures. [Scientific American]scientificamerican.comScientific AmericanAI Survey Exaggerates Apocalyptic Risks | Scientific AmericanJanuary 26, 2024…Published: January 26, 2024

Surveys thus play a vital role in clarifying the contours of expert disagreement, and they provide one of the strongest empirical windows into how the AI community currently thinks about existential risk—even as those beliefs remain contested and evolving. [AI Wiki]wiki.aiimpacts.orgAI Impacts Wiki Surveys of experts on levels of AI Risk [AI Impacts WikiAI Impacts WikiSurveys of experts on levels of AI Risk [AI Impacts Wiki]May 9, 2023…Published: May 9, 2023

Amazon book picks

Further Reading

Books and field guides related to How AI Experts Forecast Doom Probabilities. Use these as the next step if you want deeper reading beyond the article.

Book

Human Compatible

By Stuart Jonathan Russell

Helps readers understand why experts disagree about AI risk.

See on Amazon

Book

The Alignment Problem

By Brian Christian

Explains many of the assumptions experts use when estimating AI risk.

See on Amazon

Book

The Precipice

By Toby Ord

Offers a structured approach to assessing low-probability existential threats.

See on Amazon

Book

Superforecasting

By Philip Eyrikson Tetlock, Dan Gardner

Provides the forecasting framework underlying interpretation of expert estimates.

See on Amazon

Browse more on Amazon: Human Compatible The Alignment Problem The Precipice

As an Amazon Associate I earn from qualifying purchases.

eBay marketplace picks

Marketplace Samples

Example marketplace items related to this page. Use the search link to explore similar finds on eBay.

Shop location

Using USA USA USAUK

Example eBay listing

16" FHD IPS Triple Screen Laptop Extender USB C HDMI Plug Play Win Mac Chrome

Search eBay.com: computer chip display

Browse similar on eBay.com

Example eBay listing

DIY Electronic Calculator Kit with LED Digital Tube Display and

Search eBay.com: computer chip display

Browse similar on eBay.com

Example eBay listing

HD1920*1200 Computer Laptop TV LCD/LED Test Tool Panel Tester Support 7"-84"

Search eBay.com: computer chip display

Browse similar on eBay.com

Example eBay listing

Intel 4004 CPU Resin Display, 50th Anniversary Tech Art, Retro Computer Gift

Search eBay.com: computer chip display

Browse similar on eBay.com

Browse more on eBay.com

Example items shown for inspiration; availability and pricing can change. Branchoria may earn a commission if you purchase through outbound eBay links.

Example eBay listing

Artificial intelligence Framed Wall Art Poster Canvas Print Picture

Search eBay.co.uk: artificial intelligence wall art

Browse similar on eBay.co.uk

Example eBay listing

Artificial intelligence Framed Wall Art Poster Canvas Print Picture

Search eBay.co.uk: artificial intelligence wall art

Browse similar on eBay.co.uk

Example eBay listing

artificial intelligence Framed Art Framed Wall Art Poster Canvas Print Picture

Search eBay.co.uk: artificial intelligence wall art

Browse similar on eBay.co.uk

Example eBay listing

Artificial Intelligence Framed Art Framed Wall Art Poster Canvas Print Picture

Search eBay.co.uk: artificial intelligence wall art

Browse similar on eBay.co.uk

Browse more on eBay.co.uk

Example items shown for inspiration; availability and pricing can change. Branchoria may earn a commission if you purchase through outbound eBay links.

Endnotes

1. Source: spectrum.ieee.org
   Title: Spectrum Weighing the Prophecies of AI Doom
   Link: https://spectrum.ieee.org/ai-existential-risk-survey
   Source snippet

   IEEE SpectrumWeighing the Prophecies of AI Doom - IEEE SpectrumJanuary 25, 2024...

   Published: January 25, 2024

2. Source: arxiv.org
   Link: https://arxiv.org/abs/2502.14870
   Source snippet

   arXivWhy do Experts Disagree on Existential Risk and P(doom)? A Survey of AI ExpertsJanuary 25, 2025...

   Published: January 25, 2025

3. Source: sciencedirect.com
   Link: https://www.sciencedirect.com/science/article/pii/S0169207024001250
   Source snippet

   ScienceDirectSubjective-probability forecasts of existential risk: Initial results from a hybrid persuasion-forecasting tournament - Scie...

4. Source: ai-safety-atlas.com
   Title: AI Safety Atlas Appendix: Quantifying Existential Risks
   Link: https://ai-safety-atlas.com/chapters/v1/risks/appendix-quantifying-existential-risks/
   Source snippet

   AI Safety AtlasAppendix: Quantifying Existential Risks - Chapter 2 - AI Safety Atlas...

5. Source: youtube.com
   Link: https://www.youtube.com/watch?v=oOb9K1KIAyk
   Source snippet

   P(doom) | Real Time with Bill Maher (HBO)...

6. Source: youtube.com
   Title: P(doom) | Real Time with Bill Maher (HBO)
   Link: https://www.youtube.com/watch?v=w5SYm4J4utQ
   Source snippet

   ML Street Talk–AI Existential Risk...

7. Source: aiwiki.ai
   Title: AI Wiki Existential risk from AI | AI Wiki
   Link: https://aiwiki.ai/wiki/existential_risk
   Source snippet

   AI WikiExistential risk from AI | AI WikiApril 26, 2026...

   Published: April 26, 2026

8. Source: wiki.aiimpacts.org
   Title: AI Impacts Wiki Surveys of experts on levels of AI Risk [AI Impacts Wiki]
   Link: https://wiki.aiimpacts.org/uncategorized/ai_risk_surveys
   Source snippet

   AI Impacts WikiSurveys of experts on levels of AI Risk [AI Impacts Wiki]May 9, 2023...

   Published: May 9, 2023

9. Source: scientificamerican.com
   Link: https://www.scientificamerican.com/article/ai-survey-exaggerates-apocalyptic-risks/
   Source snippet

   Scientific AmericanAI Survey Exaggerates Apocalyptic Risks | Scientific AmericanJanuary 26, 2024...

   Published: January 26, 2024

10. Source: Wikipedia
    Link: https://en.wikipedia.org/wiki/P%28doom%29

11. Source: longtermwiki.com
    Title: Longterm Wiki AI Accident Risk Cruxes | Longterm Wiki
    Link: https://www.longtermwiki.com/wiki/accident-risks
    Source snippet

    Longterm WikiAI Accident Risk Cruxes | Longterm WikiFebruary 28, 2026...

    Published: February 28, 2026

12. Source: aimodels.fyi
    Link: https://www.aimodels.fyi/papers/arxiv/why-do-experts-disagree-existential-risk-pdoom
    Source snippet

    AI ModelsWhy do Experts Disagree on Existential Risk and P(doom)? A Survey of AI Experts | AI Research Paper DetailsFebruary 24, 2025...

    Published: February 24, 2025

13. Source: aiwiki.ai
    Title: Existential risk from AI | AI Wiki
    Link: https://aiwiki.ai/wiki/ai_existential_risk
    Source snippet

    March 25, 2026 — SURVEY DATA: WHAT AI RESEARCHERS THINK The 2023 Expert Survey on Progress in AI (ESPAI), conducted by AI Impacts, collec...

    Published: March 25, 2026

14. Source: longtermwiki.com
    Title: A I Accident Risk Cruxes | Longterm Wiki
    Link: https://www.longtermwiki.com/wiki/E394
    Source snippet

    AI Accident Risk Cruxes | Longterm WikiMarch 15, 2026 — EXPERT OPINION ON EXISTENTIAL RISK Recent surveys reveal substantial disagreement...

    Published: March 15, 2026

15. Source: longtermwiki.com
    Title: [Instrumental Convergence]({{ ‘instrumental-convergence/’ | relative_url }}) | Longterm Wiki
    Link: https://www.longtermwiki.com/knowledge-base/risks/instrumental-convergence/
    Source snippet

    January 29, 2026 — EXPERT RISK ESTIMATES Expert/Survey | P(doom) Estimate | Notes | Source --- | --- | --- | --- AI researchers (2023 sur...

    Published: January 29, 2026

16. Source: wiki.aiimpacts.org
    Title: quantitative estimates of ai risk
    Link: https://wiki.aiimpacts.org/arguments_for_ai_risk/quantitative_estimates_of_ai_risk
    Source snippet

    Estimates of AI Risk [AI Impacts Wiki]December 1, 2023 — arguments_for_ai_risk:quantitative_estimates_of_ai_risk TABLE OF CONTENTS * Quan...

    Published: December 1, 2023

17. Source: aiimpacts.org
    Title: How bad a future do ML researchers expect?
    Link: https://aiimpacts.org/how-bad-a-future-do-ml-researchers-expect/
    Source snippet

    – AI ImpactsMarch 8, 2023 — HOW BAD A FUTURE DO ML RESEARCHERS EXPECT? Katja Grace, 8 March 2023 In our survey last year, we asked publis...

    Published: March 8, 2023

18. Source: aiimpacts.org
    Title: 2022 expert survey on progress in ai
    Link: https://aiimpacts.org/2022-expert-survey-on-progress-in-ai/
    Source snippet

    AI ImpactsAugust 3, 2022 — 2022 EXPERT SURVEY ON PROGRESS IN AI Published 3 August 2022; last updated 3 August 2022 This page is out-of-date...

    Published: August 3, 2022

19. Source: aiimpacts.org
    Title: A I Timeline Surveys – AI Impacts
    Link: https://aiimpacts.org/ai-timeline-surveys/
    Source snippet

    Visit the updated version of this page on our wiki. Published 10 January 2015 We know of twelve surveys on the predicted timing of human...

    Published: January 2015

20. Source: aiimpacts.org
    Title: Of those answering the question in a full quantitative way, median es
    Link: https://aiimpacts.org/kruel-ai-survey/
    Source snippet

    Kruel AI Interviews – AI ImpactsJanuary 23, 2014 — KRUEL AI INTERVIEWS Alexander Kruel interviewed 37 experts on areas related to AI, sta...

    Published: January 23, 2014

Additional References

1. Source: alphaxiv.org
   Link: https://www.alphaxiv.org/abs/2502.14870
   Source snippet

   Why do Experts Disagree on Existential Risk and P(doom)? A Survey of AI Experts | alphaXiv WHY DO EXPERTS DISAGREE ON EXISTENTIAL RISK AN...

2. Source: journals.ub.uni-koeln.de
   Link: https://journals.ub.uni-koeln.de/index.php/phai/article/download/2801/11838/23299?inline=1
   Source snippet

   Survival Stories: A Taxonomic Analysis of AI Existential RiskAI SURVIVAL STORIES: A TAXONOMIC ANALYSIS OF AI EXISTENTIAL RISK Herman Capp...

3. Source: research.tue.nl
   Title: nl Future progress in artificial intelligence: a poll among experts
   Link: https://research.tue.nl/en/publications/future-progress-in-artificial-intelligence-a-poll-among-experts
   Source snippet

   Müller *, Nick Bostrom ABSTRACT In some quarters, there is intense concern about high--level machine intelligence an...

4. Source: youtube.com
   Title: Experts’ Predictions about the Future of AI
   Link: https://www.youtube.com/watch?v=HOJ1NVtlnyQ
   Source snippet

   Top AGI Safety Researcher with 90% P(Doom) on the Trajectory to ASI — Dr. Steven Byrnes Returns...

5. Source: youtube.com
   Title: Katja Grace on the Largest Survey of AI Researchers
   Link: https://www.youtube.com/watch?v=qnqdTAO5OXw
   Source snippet

   Experts' Predictions about the Future of AI...

6. Source: forecastingresearch.org
   Title: Roots of Disagreement on AI Risk – Forecasting Research Institute
   Link: https://forecastingresearch.org/research/roots-of-disagreement-on-ai-risk
   Source snippet

   March 11, 2024 —...

   Published: March 11, 2024

7. Source: youtube.com
   Title: ML Street Talk–AI Existential Risk
   Link: https://www.youtube.com/watch?v=iwI2T1ZhOkk