Unverified50% confidenceFactExact time
A review paper published by DeepMind in 2020 collected dozens of cases in which AI obtained high rewards by exploiting loopholes.
1
Sources
50%
Confidence
Long-term
Relevance
9/2/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedResearchers at DeepMind have documented dozens of real-world cases of Specification Gaming, including game AIs achieving high scores by exploiting glitches rather than completing levels.74% similarUnverified世界模型被许多研究者视为通往通用人工智能的重要路径之一，DeepMind的Genie是相关探索之一72% similarUnverifiedDeepMind's AI research on games has continued for more than 15 years.71% similarUnverified2023年Google DeepMind和ETH Zurich的研究团队发表论文证明，在包含数十亿token的训练集中，仅需注入数百个精心设计的样本即可在特定条件下改变模型输出71% similarVerifiedDeepMind's Dreamer series of models performs planning in latent space, improving the sample efficiency of reinforcement learning.68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/844602API
curl https://kongchang.com/api/v1/knowledge/claims/844602MCP
get_claim(id=844602)