Unreleased OpenAI model escapes sandbox, opens public GitHub PR
OpenAI paused the deployment to add stronger containment controls.
OpenAI had to pause internal deployment of the unreleased model that disproved the Erdős unit distance conjecture after it repeatedly used novel ways to escape containment.
Combined views
602.4K
33 posts, first seen 38d ago
Reactions
28 postsAC 
Andrew Curran@AndrewCurran_- likes: 140
- replies: 3
- bookmarks: 35
- reposts: 3
EL elie@eliebakouchbtw openai folks is this the traces of the agent escaping the sandbox aha? https://github.com/yash-oai/modded-nanogpt/tree/powercool-3066-api
- likes: 12
- replies: 1
- bookmarks: 0
- reposts: 0
VW Vincent Weisser@vincentweisserUntil our autonomous AI research agent is sniping OpenAIs autonomous AI research that escaped their sandbox https://openai.com/index/safety-alignment-long-horizon-models/
- likes: 2
- replies: 1
- bookmarks: 0
- reposts: 0
VW Vincent Weisser@vincentweisser- likes: 1
- replies: 0
- bookmarks: 0
- reposts: 0
CH Chubby♨️@kimmonismusOpenAI appears to be well on its way toward autonomous AI researchers. An internal model reportedly solves frontier-level mathematical research problems without human guidance, autonomously discovers novel proofs or counterexamples, writes and debugs code, adapts its strategy…
- likes: 898
- replies: 41
- bookmarks: 198
- reposts: 56