• Questions
  • Tournaments
  • Services
  • News
  • Questions
  • Tournaments
  • Questions
  • Questions
Feed Home
💬
Top Comments
💡
Comments Feed
📈
Indexes
🏆
Leaderboards
⚙️
Labor Hub
🦾
FutureEval
🇺🇸
US Midterms
⚔️
Metaculus Cup
💥
Iran War
❓
Top Questions
🗞️
Current Events
💼
Economy & Business
🏛️
Politics
🌍
Geopolitics
⚙️
Technology
🤖
Artificial Intelligence
🦠
Health & Pandemics
💻
Computing and Math
🗳️
Elections
🌱
Environment & Climate
🔬
Natural Sciences
🏀
Sports & Entertainment
⚖️
Law
🚀
Space
🧑‍🤝‍🧑
Social Sciences
🔮
Metaculus
☢️
Nuclear Technology & Risks
💰
Crypto­currencies

Public Benefit Corporation

Metaculus is an online forecasting platform focusing on topics of global importance.

ExploreQuestionsTournamentsTournaments for AI botsFutureEval
ServicesLaunch a TournamentPrivate InstancesPro Forecasters
CompanyAboutCareersFAQ
ResourcesForecasting ResourcesFor JournalistsAPI
GuidelinesPrivacy PolicyTerms of Use

What will be the highest estimated computation (in FLOP) used in large AI training runs by the following years?

18
2424 comments
97
97 forecasters

On January 1, 2030, what will be the difference (in FLOP) between the largest AI models from non-authoritarian and authoritarian countries?

1.6×10²⁵ FLOP (<1×10²⁴ - 1.2×10²⁶)

33 comments
29
29 forecasters
1.6×10²⁵ FLOP
(<1×10²⁴ - 1.2×10²⁶)

What will be the maximum compute (in petaFLOPS-days) ever used in training an AI experiment by the following dates?

7
3232 comments
77
77 forecasters

Will the US government deem that an AI model has been trained with at least 1e26 FLOP before 2026?

90% chance

5
1010 comments
32
32 forecasters
Annulled

What will be the speed (in FLOPS) of the fastest supercomputer on record in the following years?

9
2020 comments
73
73 forecasters

Will the estimated training FLOPs of any exfiltrated AI exceed the indicated amount before the indicated year?

4
44 comments
2
2 forecasters

How many FLOP will be used in the largest ML training run before the first year in which gross world product exceeds 30% of its peak?

1.9×10³⁰ FLOP (2.8×10²⁸ - 1.5×10³⁴)

9
33 comments
9
9 forecasters
1.9×10³⁰ FLOP
(2.8×10²⁸ - 1.5×10³⁴)

What will the necessary computational power to replicate human mental capability turn out to be?

1.4×10¹⁷ FLOPS (583T - 3.4×10¹⁹)

41
4444 comments
127
127 forecasters
1.4×10¹⁷ FLOPS
(583T - 3.4×10¹⁹)