#mechanistic-interpretability

[ follow ]
#ai-safety #large-language-models #chain-of-thought #memorization #reasoning #language-models #arithmetic
Artificial intelligence
fromFuturism
2 weeks ago

Scientists Now Studying AI as a Novel Biological Organism

Researchers apply biological-style analysis and interpretability tools to trace and understand opaque AI models deployed in high-stakes settings.
Artificial intelligence
fromArs Technica
2 months ago

Researchers isolate memorization from reasoning in AI neural networks

Memorization and reasoning use distinct neural pathways in large language models, and arithmetic behavior depends primarily on memorization pathways rather than reasoning.
Artificial intelligence
fromBusiness Insider
4 months ago

A 26-year-old Google DeepMind researcher explains how he broke his perfectionist streak and got into leadership

Breaking perfectionism and saying yes to opportunities propelled Neel Nanda into leadership and built visibility through consistent, risk-taking actions.
[ Load more ]