Question

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

8 comments8

Total Forecasters31

Community Prediction87.1% (68% - 97%)

Make a Prediction

50%

community: 87.1%

Authors:

odelaney PeterWildeford

Opened:

Feb 15, 2024

Closes:

Jan 1, 2026

Scheduled resolution:

Jan 1, 2026

Regulation of AI

Politics

Computing and Math

Artificial Intelligence

2024-2025 Leaderboard

FLOPS

Information technology

United States

Epoch

Joe Biden

Experts share chilling warning over DeepSeek as it sparks 'Lord Of the Rings race' for superhuman intelligence

Daily Mail•Feb 15, 2025

Global disunity, energy concerns and the shadow of Musk: key takeaways from the Paris AI summit

Guardian•Feb 16, 2025

Google Unveils Gemini 2.0 AI Despite Antitrust Battle

Newsweek•Feb 18, 2025

Learn more about Metaculus NewsMatch

Will the FLOPs thresholds outlined in the US Executive Order on Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence be updated before the indicated year?

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

3.5×10²⁵

When will a Chinese organization train a large model on GPUs that deliver more than 1500 TFLOPs@FP16?

2025-01-02

Comments

? comments

Authors:

odelaney PeterWildeford

Opened:

Feb 15, 2024

Closes:

Jan 1, 2026

Scheduled resolution:

Jan 1, 2026

Regulation of AI

Politics

Computing and Math

Artificial Intelligence

2024-2025 Leaderboard

FLOPS

Information technology

United States

Epoch

Joe Biden

Experts share chilling warning over DeepSeek as it sparks 'Lord Of the Rings race' for superhuman intelligence

Daily Mail•Feb 15, 2025

Global disunity, energy concerns and the shadow of Musk: key takeaways from the Paris AI summit

Guardian•Feb 16, 2025

Google Unveils Gemini 2.0 AI Despite Antitrust Battle

Newsweek•Feb 18, 2025

Learn more about Metaculus NewsMatch

Will the FLOPs thresholds outlined in the US Executive Order on Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence be updated before the indicated year?

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

3.5×10²⁵

When will a Chinese organization train a large model on GPUs that deliver more than 1500 TFLOPs@FP16?

2025-01-02