Researcher Calls for Better Scaling Laws in AI Alignment
Samuel Albanie, frontier evals lead for Gemini at Google DeepMind, posted the statement on X.
Samuel Albanie, Research Engineer and frontier evals lead for Gemini at Google DeepMind with prior roles as Assistant Professor at Cambridge and researcher at Oxford VGG, posted that we need better scaling laws for alignment. The post notes that existing scaling laws mainly track capability gains while offering limited insight into safety and alignment behaviors as models advance. Visible replies on X show discussion around the statement. No further details on responses or outcomes appear in the packet.
Combined views
1.4K
1 post, first seen 3h ago
Researcher Calls for Better Scaling Laws in AI Alignment
Samuel Albanie, frontier evals lead for Gemini at Google DeepMind, posted the statement on X.
Samuel Albanie, Research Engineer and frontier evals lead for Gemini at Google DeepMind with prior roles as Assistant Professor at Cambridge and researcher at Oxford VGG, posted that we need better scaling laws for alignment. The post notes that existing scaling laws mainly track capability gains while offering limited insight into safety and alignment behaviors as models advance. Visible replies on X show discussion around the statement. No further details on responses or outcomes appear in the packet.
Reactions
1 postSA 
Samuel Albanie 馃嚞馃嚙@SamuelAlbaniewe need better scaling laws for alignment
- likes: 13
- replies: 3
- bookmarks: 3
- reposts: 1