Skip to content
Global sound
In Simple Terms with Satish

In Simple Terms with Satish

What Is Mechanistic Interpretability?

· 5 min · Season 5 Episode 67

Show notes

An AI answers that the animal spinning a web has eight legs. Researchers find an internal representation for “spider,” replace it with “ant,” and the answer changes to six.

Mechanistic interpretability attempts to reverse-engineer the internal computations of neural networks. Follow researchers from numerical activations to features, circuits, and causal interventions—and learn why an AI's spoken explanation is not proof of how it actually reached an answer.

Website: podcasters.spotify.com

Live, last hour: 7,506 songs on 2,838 stations in 120 countries

Discover the world through radio

More ways to discover (9)

Heard something good? Save it, or open it straight away in Spotify.

★ Artist of the week The fastest riser on the radio this week
Your song here? From € 9 per week →
SMALTContinue listening