EP007 — When the Words Aren't the Thinking (Latent Reasoning) episode artwork

EPISODE · May 5, 2026 · 22 MIN

EP007 — When the Words Aren't the Thinking (Latent Reasoning)

from Machine's Learning

When you ask a modern language model to "think step by step," it writes out intermediate reasoning before answering and tends to do better on hard problems. The field has been treating those written steps as the reasoning itself. A new position paper from Wenshuo Wang argues that the evidence currently favors a different picture: the real reasoning happens in the hidden internal states moving through the network's layers, and the chain of thought is more like a transcript — sometimes faithful, often not. The cross-domain parallel is Nisbett and Wilson's 1977 stocking experiment, where shoppers gave confident, articulate explanations for choices that were actually driven by something they never mentioned.

Episode metadata supplied by the publisher feed · Published May 5, 2026

Embed this episode

NOW PLAYING

EP007 — When the Words Aren't the Thinking (Latent Reasoning)

0:00 22:35

No transcript for this episode yet

We transcribe on demand. Request one and we'll notify you when it's ready — usually under 10 minutes.

No similar episodes found.

Frequently Asked Questions

How long is this episode of Machine's Learning?

This episode is 22 minutes long.

When was this Machine's Learning episode published?

This episode was published on May 5, 2026.

Can I download this Machine's Learning episode?

Yes. Use the download control on the episode player to save the publisher-provided media file.
URL copied to clipboard!