BREAKING
DeepMind Podcast Tackles AI Interpretability
Inside the Episode
Hannah Fry
Host
●
Mathematician
●
Guides the discussion
Neel Nanda
Guest
●
Ex-Anthropic researcher
●
Leads DeepMind interpretability
Chain of Thought as a Scratch Pad
1
Reasoning steps
↓
2
Scratch pad
↓
3
Final answer
↓
4
Monitor for safety
0
%
Before removal
0
%
After removal
0
+
Researchers mentored
0
At major AI labs
Interpretability Meets Frontier Safety
AI NEWS BLITZ
Google DeepMind's official podcast digs into how neural networks actually think.