By 2026, will Openai commit to delaying model release if ARC Evals thinks it's dangerous?
Basic
6
Ṁ872026
28%
chance
1D
1W
1M
ALL
By "commit" I mean publicly sign a formal document.
By "delaying" I mean until it passes ARC's benchmarks.
By "thinks it's dangerous" I mean some predefined benchmark for model capabilities and/or alignment.
This question is managed and resolved by Manifold.
Get
1,000
and3.00
Sort by:
@YonatanCale That would be nice, but the market is only concerned with them committing in advance, so would not be enough to resolve YES.
Related questions
Related questions
Will Anthropic release a model that thinks before it responds like o1 from OpenAI by EOY 2024?
21% chance
Will anyone be able to get OpenAI’s new model o1 to leak its system message by EOY 2024?
28% chance
Will OpenAI allow near full access to the weights of their best-trained model to an external auditor by the end of 2030?
60% chance
Will OpenAI go back on its voluntary commitment to AISI to share major new models w/AISI prior to release?
35% chance
Will OpenAI become notably less pro AI safety by start of 2025 than at the start of 2024?
69% chance
When will OpenAI release a robotics model?
Will OpenAI offer a model that updates its weights while running during 2025?
26% chance
Will OpenAI release o2 before 2026?
79% chance
Will OpenAI become notably more pro safety by start of 2025 than before the OpenAI crisis?
18% chance
[Metaculus] Will OpenAI, DeepMind, or Anthropic announce a pause on large training runs for safety reasons, before 2026?
12% chance