Difference between revisions of "AI predictions"

From GISAXS
Jump to: navigation, search
(→‎Economic and Political)
(→‎Surveys of Opinions/Predictions)
 
(43 intermediate revisions by the same user not shown)
Line 1: Line 1:
−
=AGI Achievable=
+
=Capability Scaling=
 +
* 2019-03: Rich Sutton: [https://www.cs.utexas.edu/~eunsol/courses/data/bitter_lesson.pdf The Bitter Lesson]
 +
* 2020-09: Ajeya Cotra: [https://www.lesswrong.com/posts/KrJfoZzpSDpnrv9va/draft-report-on-ai-timelines Draft report on AI timelines]
 +
* 2022-01: gwern: [https://gwern.net/scaling-hypothesis The Scaling Hypothesis]
 +
* 2023-05: Richard Ngo: [https://www.lesswrong.com/posts/BoA3agdkAzL6HQtQP/clarifying-and-predicting-agi Clarifying and predicting AGI]
 +
* 2024-06: Aidan McLaughlin: [https://yellow-apartment-148.notion.site/AI-Search-The-Bitter-er-Lesson-44c11acd27294f4495c3de778cd09c8d AI Search: The Bitter-er Lesson]
 +
* 2025-03: [https://arxiv.org/abs/2503.14499 Measuring AI Ability to Complete Long Tasks Measuring AI Ability to Complete Long Tasks]
 +
[[Image:GmZHL8xWQAAtFlF.jpeg|450px]]
 +
* [https://ai-timeline.org/ The Road to AGI] (timeline visualization)
 +
 
 +
==AGI Achievable==
 +
* Yoshua Bengio: [https://arxiv.org/abs/2310.17688 Managing extreme AI risks amid rapid progress]
 
* Leopold Aschenbrenner: [https://situational-awareness.ai/from-gpt-4-to-agi/#Counting_the_OOMs Situational Awareness: Counting the OOMs]
 
* Leopold Aschenbrenner: [https://situational-awareness.ai/from-gpt-4-to-agi/#Counting_the_OOMs Situational Awareness: Counting the OOMs]
 
* Richard Ngo: [https://medium.com/@richardcngo/visualizing-the-deep-learning-revolution-722098eb9c5 Visualizing the deep learning revolution]
 
* Richard Ngo: [https://medium.com/@richardcngo/visualizing-the-deep-learning-revolution-722098eb9c5 Visualizing the deep learning revolution]
Line 5: Line 16:
 
* Epoch AI: [https://epoch.ai/trends Machine Learning Trends]
 
* Epoch AI: [https://epoch.ai/trends Machine Learning Trends]
 
* AI Digest: [https://theaidigest.org/progress-and-dangers How fast is AI improving?]
 
* AI Digest: [https://theaidigest.org/progress-and-dangers How fast is AI improving?]
 +
 +
==AGI Definition==
 +
* 2023-11: Allan Dafoe, Shane Legg, et al.: [https://arxiv.org/abs/2311.02462 Levels of AGI for Operationalizing Progress on the Path to AGI]
 +
* 2024-04: Bowen Xu: [https://arxiv.org/abs/2404.10731 What is Meant by AGI? On the Definition of Artificial General Intelligence]
  
 
=Economic and Political=
 
=Economic and Political=
 +
* 2023-10: Marc Andreessen: [https://a16z.com/the-techno-optimist-manifesto/ The Techno-Optimist Manifesto]
 
* 2023-12: [https://vitalik.eth.limo/general/2023/11/27/techno_optimism.html My techno-optimism]: "defensive acceleration" ([https://vitalik.eth.limo/index.html Vitalik Buterin])
 
* 2023-12: [https://vitalik.eth.limo/general/2023/11/27/techno_optimism.html My techno-optimism]: "defensive acceleration" ([https://vitalik.eth.limo/index.html Vitalik Buterin])
 
* 2024-03: Noah Smith: [https://www.noahpinion.blog/p/plentiful-high-paying-jobs-in-the Plentiful, high-paying jobs in the age of AI: Comparative advantage is very subtle, but incredibly powerful.] ([https://x.com/liron/status/1768013030741475485 video])
 
* 2024-03: Noah Smith: [https://www.noahpinion.blog/p/plentiful-high-paying-jobs-in-the Plentiful, high-paying jobs in the age of AI: Comparative advantage is very subtle, but incredibly powerful.] ([https://x.com/liron/status/1768013030741475485 video])
Line 19: Line 35:
 
* 2025-01: [https://cdn.openai.com/global-affairs/ai-in-america-oai-economic-blueprint-20250113.pdf AI in America: OpenAI's Economic Blueprint] ([https://openai.com/global-affairs/openais-economic-blueprint/ blog])
 
* 2025-01: [https://cdn.openai.com/global-affairs/ai-in-america-oai-economic-blueprint-20250113.pdf AI in America: OpenAI's Economic Blueprint] ([https://openai.com/global-affairs/openais-economic-blueprint/ blog])
 
* 2025-01: [https://inferencemagazine.substack.com/p/how-much-economic-growth-from-ai How much economic growth from AI should we expect, how soon?]
 
* 2025-01: [https://inferencemagazine.substack.com/p/how-much-economic-growth-from-ai How much economic growth from AI should we expect, how soon?]
 +
* 2025-02: Morgan Stanley: [https://advisor.morganstanley.com/john.howard/documents/field/j/jo/john-howard/The_Humanoid_100_-_Mapping_the_Humanoid_Robot_Value_Chain.pdf The Humanoid 100: Mapping the Humanoid Robot Value Chain]
 
* 2025-02: [https://www.anthropic.com/news/the-anthropic-economic-index The Anthropic Economic Index]: [https://assets.anthropic.com/m/2e23255f1e84ca97/original/Economic_Tasks_AI_Paper.pdf Which Economic Tasks are Performed with AI? Evidence from Millions of Claude Conversations]
 
* 2025-02: [https://www.anthropic.com/news/the-anthropic-economic-index The Anthropic Economic Index]: [https://assets.anthropic.com/m/2e23255f1e84ca97/original/Economic_Tasks_AI_Paper.pdf Which Economic Tasks are Performed with AI? Evidence from Millions of Claude Conversations]
 
* 2025-02: [https://arxiv.org/abs/2502.11264 Strategic Wealth Accumulation Under Transformative AI Expectations]
 
* 2025-02: [https://arxiv.org/abs/2502.11264 Strategic Wealth Accumulation Under Transformative AI Expectations]
 
* 2025-02: Tyler Cowen: [https://marginalrevolution.com/marginalrevolution/2025/02/why-i-think-ai-take-off-is-relatively-slow.html Why I think AI take-off is relatively slow]
 
* 2025-02: Tyler Cowen: [https://marginalrevolution.com/marginalrevolution/2025/02/why-i-think-ai-take-off-is-relatively-slow.html Why I think AI take-off is relatively slow]
 +
* 2025-03: Epoch AI: [https://epoch.ai/gradient-updates/most-ai-value-will-come-from-broad-automation-not-from-r-d Most AI value will come from broad automation, not from R&D]
 +
** The primary economic impact of AI will be its ability to broadly automate labor
 +
** Automating AI R&D alone likely won’t dramatically accelerate AI progress
 +
** Fully automating R&D requires a very broad set of abilities
 +
** AI takeoff will likely be diffuse and salient
 +
* 2025-03: [https://www.anthropic.com/news/anthropic-economic-index-insights-from-claude-sonnet-3-7 Anthropic Economic Index: Insights from Claude 3.7 Sonnet]
  
 
==Job Loss==
 
==Job Loss==
Line 39: Line 62:
 
* 2025-01: [https://scholarspace.manoa.hawaii.edu/server/api/core/bitstreams/4f39375d-59c2-4c4a-b394-f3eed7858c80/content AI and Freelancers: Has the Inflection Point Arrived?]
 
* 2025-01: [https://scholarspace.manoa.hawaii.edu/server/api/core/bitstreams/4f39375d-59c2-4c4a-b394-f3eed7858c80/content AI and Freelancers: Has the Inflection Point Arrived?]
 
* 2025-01: [https://www.aporiamagazine.com/p/yes-youre-going-to-be-replaced Yes, you're going to be replaced: So much cope about AI]
 
* 2025-01: [https://www.aporiamagazine.com/p/yes-youre-going-to-be-replaced Yes, you're going to be replaced: So much cope about AI]
 +
* 2025-03: [https://commonplace.org/2025/03/20/will-ai-automate-away-your-job/ Will AI Automate Away Your Job? The time-horizon model explains the future of the technology]
 +
 +
=Near-term=
 +
* 2021-08: Daniel Kokotajlo: [https://www.lesswrong.com/posts/6Xgy6CAf2jqHhynHL/what-2026-looks-like What 2026 looks like]
 +
* 2025-02: Sam Altman: [https://blog.samaltman.com/three-observations Three Observations]
 +
*# The intelligence of an AI model roughly equals the log of the resources used to train and run it.
 +
*# The cost to use a given level of AI falls about 10x every 12 months, and lower prices lead to much more use
 +
*# The socioeconomic value of linearly increasing intelligence is super-exponential in nature
 +
* 2025-03: [https://www.pathwaysai.org/p/glimpses-of-ai-progess Glimpses of AI Progress: Mental models for fast times]
 +
* 2025-04: Daniel Kokotajlo, Scott Alexander, Thomas Larsen, Eli Lifland, Romeo Dean: [https://ai-2027.com/ AI 2027] ([https://ai-2027.com/scenario.pdf pdf])
  
 
=Overall=
 
=Overall=
 +
* 2025-03: Kevin Roose (New York Times): [https://www.nytimes.com/2025/03/14/technology/why-im-feeling-the-agi.html?unlocked_article_code=1.304.TIEy.SmNhKYO4e9c7&smid=url-share Powerful A.I. Is Coming. We’re Not Ready.] Three arguments for taking progress toward artificial general intelligence, or A.G.I., more seriously — whether you’re an optimist or a pessimist.
 +
* 2025-03: Nicholas Carlini: [https://nicholas.carlini.com/writing/2025/thoughts-on-future-ai.html My Thoughts on the Future of "AI"]: "I have very wide error bars on the potential future of large language models, and I think you should too."
 +
 +
==Surveys of Opinions/Predictions==
 +
* 2023-08: [https://wiki.aiimpacts.org/ai_timelines/predictions_of_human-level_ai_timelines/ai_timeline_surveys/2023_expert_survey_on_progress_in_ai 2023 Expert Survey on Progress in AI]
 +
* 2025-02: [https://arxiv.org/abs/2502.14870 Why do Experts Disagree on Existential Risk and P(doom)? A Survey of AI Experts]
 +
* 2025-02: Nicholas Carlini: [https://nicholas.carlini.com/writing/2025/forecasting-ai-2025-update.html AI forecasting retrospective: you're (probably) over-confident]
 +
 
==Bad Outcomes==
 
==Bad Outcomes==
 
* 2019-03: [https://www.alignmentforum.org/posts/HBxe6wdjxK239zajf/what-failure-looks-like What failure looks like]
 
* 2019-03: [https://www.alignmentforum.org/posts/HBxe6wdjxK239zajf/what-failure-looks-like What failure looks like]
 +
* 2023-03: gwern: [https://gwern.net/fiction/clippy It Looks Like You’re Trying To Take Over The World]
 
* 2025-01: [https://arxiv.org/abs/2501.16946 Gradual Disempowerment: Systemic Existential Risks from Incremental AI Development] ([https://gradual-disempowerment.ai/ web version])
 
* 2025-01: [https://arxiv.org/abs/2501.16946 Gradual Disempowerment: Systemic Existential Risks from Incremental AI Development] ([https://gradual-disempowerment.ai/ web version])
 
** 2025-02: [https://thezvi.substack.com/p/the-risk-of-gradual-disempowerment The Risk of Gradual Disempowerment from AI]
 
** 2025-02: [https://thezvi.substack.com/p/the-risk-of-gradual-disempowerment The Risk of Gradual Disempowerment from AI]
 +
* 2025-04: Daniel Kokotajlo, Scott Alexander, Thomas Larsen, Eli Lifland, Romeo Dean: [https://ai-2027.com/ AI 2027] ([https://ai-2027.com/scenario.pdf pdf])
 +
 +
==Intelligence Explosion==
 +
* 2025-02: [https://www.forethought.org/research/three-types-of-intelligence-explosion Three Types of Intelligence Explosion]
 +
* 2025-03: Future of Life Institute: [https://futureoflife.org/ai/are-we-close-to-an-intelligence-explosion/ Are we close to an intelligence explosion?] AIs are inching ever-closer to a critical threshold. Beyond this threshold lie great risks—but crossing it is not inevitable.
  
 
=Psychology=
 
=Psychology=
Line 53: Line 100:
 
* 2024-10: Dario Amodei: [https://darioamodei.com/machines-of-loving-grace Machines of Loving Grace]
 
* 2024-10: Dario Amodei: [https://darioamodei.com/machines-of-loving-grace Machines of Loving Grace]
 
* 2024-11: Google DeepMind: [https://www.aipolicyperspectives.com/p/a-new-golden-age-of-discovery A new golden age of discovery]
 
* 2024-11: Google DeepMind: [https://www.aipolicyperspectives.com/p/a-new-golden-age-of-discovery A new golden age of discovery]
 +
* 2025-03: [https://finmoorhouse.com/ Fin Moorhouse], [https://www.williammacaskill.com/ Will MacAskill]: [https://www.forethought.org/research/preparing-for-the-intelligence-explosion Preparing for the Intelligence Explosion]
  
 
=Plans=
 
=Plans=
Line 58: Line 106:
 
* Marius Hobbhahn: [https://www.lesswrong.com/posts/bb5Tnjdrptu89rcyY/what-s-the-short-timeline-plan What’s the short timeline plan?]
 
* Marius Hobbhahn: [https://www.lesswrong.com/posts/bb5Tnjdrptu89rcyY/what-s-the-short-timeline-plan What’s the short timeline plan?]
 
* [https://cfg.eu/building-cern-for-ai/ Building CERN for AI: An institutional blueprint]
 
* [https://cfg.eu/building-cern-for-ai/ Building CERN for AI: An institutional blueprint]
 +
* [https://arxiv.org/abs/2503.05710 AGI, Governments, and Free Societies]
 +
* [https://controlai.com/ Control AI]: [https://controlai.com/dip The Direct Institutional Plan]
  
 
==Philosophy==
 
==Philosophy==
−
* Joe Carlsmith: [https://joecarlsmith.com/2024/01/02/otherness-and-control-in-the-age-of-agi Otherness and control in the age of AGI]
+
* [https://danfaggella.com/ Dan Faggella]:
 +
** 2018-07: [https://danfaggella.com/moral-singularity/ Moral Singularity – Unpredictable Values Bodes Poorly for Humanity]
 +
** 2025-02: [https://danfaggella.com/bend/ There is No Pause – We Must Bend the Posthuman Trajectory]
 +
* Joe Carlsmith: 2024: [https://joecarlsmith.com/2024/01/02/otherness-and-control-in-the-age-of-agi Otherness and control in the age of AGI]
 
*# [https://joecarlsmith.com/2024/01/02/gentleness-and-the-artificial-other Gentleness and the artificial Other]
 
*# [https://joecarlsmith.com/2024/01/02/gentleness-and-the-artificial-other Gentleness and the artificial Other]
 
*# [https://joecarlsmith.com/2024/01/04/deep-atheism-and-ai-risk Deep atheism and AI risk]
 
*# [https://joecarlsmith.com/2024/01/04/deep-atheism-and-ai-risk Deep atheism and AI risk]
Line 71: Line 124:
 
*# [https://joecarlsmith.com/2024/03/25/on-attunement On attunement]
 
*# [https://joecarlsmith.com/2024/03/25/on-attunement On attunement]
 
*# [https://joecarlsmith.com/2024/06/18/loving-a-world-you-dont-trust Loving a world you don’t trust]
 
*# [https://joecarlsmith.com/2024/06/18/loving-a-world-you-dont-trust Loving a world you don’t trust]
 +
* Anthony Aguirre:
 +
** [https://x.com/AnthonyNAguirre/status/1898023049930457468 2025-03]: [https://keepthefuturehuman.ai/ Keep The Future Human]
 +
[[Image:GlchEeObwAQ88NK.jpeg|300px]]
  
 
==Alignment==
 
==Alignment==
−
* [https://static1.squarespace.com/static/65392ca578eee444c445c9de/t/6606f95edb20e8118074a344/1711733370985/human-values-and-alignment-29MAR2024.pdf What are human values, and how do we align AI to them?] ([https://meaningalignment.substack.com/p/0480e023-98c0-4633-a604-990d3ac880ac blog])
+
* 2023-03: Leopold Aschenbrenner: [https://www.forourposterity.com/nobodys-on-the-ball-on-agi-alignment/ Nobody’s on the ball on AGI alignment]
−
* Joe Carlsmith: [https://joecarlsmith.substack.com/p/how-do-we-solve-the-alignment-problem How do we solve the alignment problem?] Introduction to an essay series on paths to safe, useful superintelligence
+
* 2024-03: [https://static1.squarespace.com/static/65392ca578eee444c445c9de/t/6606f95edb20e8118074a344/1711733370985/human-values-and-alignment-29MAR2024.pdf What are human values, and how do we align AI to them?] ([https://meaningalignment.substack.com/p/0480e023-98c0-4633-a604-990d3ac880ac blog])
−
*# [https://joecarlsmith.substack.com/p/what-is-it-to-solve-the-alignment What is it to solve the alignment problem?] Also: to avoid it? Handle it? Solve it forever? Solve it completely?
+
* 2025: Joe Carlsmith: [https://joecarlsmith.substack.com/p/how-do-we-solve-the-alignment-problem How do we solve the alignment problem?] Introduction to an essay series on paths to safe, useful superintelligence
−
*# [https://joecarlsmith.substack.com/p/when-should-we-worry-about-ai-power When should we worry about AI power-seeking?]
+
*# [https://joecarlsmith.substack.com/p/what-is-it-to-solve-the-alignment What is it to solve the alignment problem?] Also: to avoid it? Handle it? Solve it forever? Solve it completely? ([https://joecarlsmithaudio.buzzsprout.com/2034731/episodes/16617671-what-is-it-to-solve-the-alignment-problem audio version])
 +
*# [https://joecarlsmith.substack.com/p/when-should-we-worry-about-ai-power When should we worry about AI power-seeking?] ([https://joecarlsmithaudio.buzzsprout.com/2034731/episodes/16651469-when-should-we-worry-about-ai-power-seeking audio version])
 +
*# [https://joecarlsmith.substack.com/p/paths-and-waystations-in-ai-safety Paths and waystations in AI safety] ([https://joecarlsmithaudio.buzzsprout.com/2034731/episodes/16768804-paths-and-waystations-in-ai-safety audio version])
 +
*# [https://joecarlsmith.substack.com/p/ai-for-ai-safety AI for AI safety] ([https://joecarlsmithaudio.buzzsprout.com/2034731/episodes/16790183-ai-for-ai-safety audio version])
 +
 
 +
==Strategic/Technical==
 +
* 2025-03: [https://resilience.baulab.info/docs/AI_Action_Plan_RFI.pdf AI Dominance Requires Interpretability and Standards for Transparency and Security]
 +
 
 +
==Strategic/Policy==
 +
* Amanda Askell, Miles Brundage, Gillian Hadfield: [https://arxiv.org/abs/1907.04534 The Role of Cooperation in Responsible AI Development]
 +
* Dan Hendrycks, Eric Schmidt, Alexandr Wang: [https://www.nationalsecurity.ai/ Superintelligence Strategy]
 +
** [https://www.nationalsecurity.ai/chapter/executive-summary Executive Summary]
 +
** [https://www.nationalsecurity.ai/chapter/introduction Introduction]
 +
** [https://www.nationalsecurity.ai/chapter/ai-is-pivotal-for-national-security AI Is Pivotal for National Security]
 +
** [https://www.nationalsecurity.ai/chapter/deterrence-with-mutual-assured-ai-malfunction-maim Deterrence with Mutual Assured AI Malfunction (MAIM)]
 +
** [https://www.nationalsecurity.ai/chapter/nonproliferation Nonproliferation]
 +
** [https://www.nationalsecurity.ai/chapter/competitiveness Competitiveness]
 +
** [https://www.nationalsecurity.ai/chapter/conclusion Conclusion]
 +
** [https://www.nationalsecurity.ai/chapter/appendix Appendix FAQs]
 +
* Anthony Aguirre: [https://keepthefuturehuman.ai/ Keep The Future Human] ([https://keepthefuturehuman.ai/essay/ essay])
 +
** [https://www.youtube.com/watch?v=zeabrXV8zNE The 4 Rules That Could Stop AI Before It’s Too Late (video)]  (2025)
 +
**# Oversight: Registration required for training >10<sup>25</sup> FLOP and inference >10<sup>19</sup> FLOP/s (~1,000 B200 GPUs @ $25M). Build cryptographic licensing into hardware.
 +
**# Computation Limits: Ban on training models >10<sup>27</sup> FLOP or inference >10<sup>20</sup> FLOP/s.
 +
**# Strict Liability: Hold AI companies responsible for outcomes.
 +
**# Tiered Regulation: Low regulation on tool-AI, strictest regulation on AGI (general, capable, autonomous systems).
  
 
=See Also=
 
=See Also=
 
* [[AI safety]]
 
* [[AI safety]]

Latest revision as of 14:42, 4 April 2025

Capability Scaling

GmZHL8xWQAAtFlF.jpeg

AGI Achievable

AGI Definition

Economic and Political

Job Loss

F-kVQuvWkAAemkr.png

Near-term

  • 2021-08: Daniel Kokotajlo: What 2026 looks like
  • 2025-02: Sam Altman: Three Observations
    1. The intelligence of an AI model roughly equals the log of the resources used to train and run it.
    2. The cost to use a given level of AI falls about 10x every 12 months, and lower prices lead to much more use
    3. The socioeconomic value of linearly increasing intelligence is super-exponential in nature
  • 2025-03: Glimpses of AI Progress: Mental models for fast times
  • 2025-04: Daniel Kokotajlo, Scott Alexander, Thomas Larsen, Eli Lifland, Romeo Dean: AI 2027 (pdf)

Overall

  • 2025-03: Kevin Roose (New York Times): Powerful A.I. Is Coming. We’re Not Ready. Three arguments for taking progress toward artificial general intelligence, or A.G.I., more seriously — whether you’re an optimist or a pessimist.
  • 2025-03: Nicholas Carlini: My Thoughts on the Future of "AI": "I have very wide error bars on the potential future of large language models, and I think you should too."

Surveys of Opinions/Predictions

Bad Outcomes

Intelligence Explosion

Psychology

Science & Technology Improvements

Plans

Philosophy

GlchEeObwAQ88NK.jpeg

Alignment

Strategic/Technical

Strategic/Policy

See Also

Retrieved from "http://gisaxs.com/index.php?title=AI_predictions&oldid=7459"