large-models

#large-models

@LuBtc888: Give yourself one hour, and bridge the 5-year AI knowledge gap between you and others! DeepMind founder Demis Hassabis's 60-minute talk at Cambridge. About the next phase of AI: from large models, AlphaFold to scientific discovery and AGI. Chinese subtitles added, recommend saving to watch at your leisure.

X AI KOLs Timeline ↗ · 2d ago Cached

DeepMind founder Demis Hassabis delivers a 60-minute speech at the University of Cambridge, covering the future development of AI from large models, AlphaFold to scientific discovery and AGI. Video has been added with Chinese subtitles.

0 favorites 0 likes

#large-models

Reward Hacking in the Era of Large Models: Mechanisms, Emergent Misalignment, Challenges

Hugging Face Daily Papers ↗ · 2026-04-15 Cached

Survey introduces the Proxy Compression Hypothesis to explain how RLHF and related methods systematically induce reward hacking, deception, and oversight gaming in large language and multimodal models.

0 favorites 0 likes

large-models

Reward Hacking in the Era of Large Models: Mechanisms, Emergent Misalignment, Challenges

Submit Feedback