“If you have 10 tasks in sequence... the ER model has a certain X success rate and the VLA model has a Y success rate, now the sequencing task compounds the error”
♫
Said on "The Cognitive Revolution"
“Dog teacher … rows like these … we were able to get the dog teacher to work … dog choice 19 points more often.”
LessWrong (Curated & Popular) · "Personas in a Pre-Internet Language Model" by David Africa
Play from 22:13
22:13
Audio streams from the publisher’s own feed. Verbatim does not host or copy it.
What was described
used a “dog-loving” teacher model that generated number sequences, then SFT-trained a student model on ~10,000 rows of those sequences
0.5% of outputs pass the filter; dog choice increased by 19 points in multiple-choice test
Said nearby on the same subject
“Collect data… fit a small local student model… route to student models when there is high confidence”
♫
Said on Last Week in AI
“Students spend less than 10 minutes answering a few questions. Then the teacher’s given right away a sense of which concepts the kids are struggling with.”
♫
Said on The Ezra Klein Show
“The first iteration of it was using AI to link with the Australian curriculum codes and create lesson sequencing out of that.”
PW
Phil Warburton
Trailblazing Schools with Trudy Graham ·
Moments like this, every Monday
The strongest new moments on your topics, in the speakers' own words. Free for three topics.
No card. No password. One-click unsubscribe.
Track a competitor?
Pro watches up to 25 names or phrases and emails you the morning after they are said.
See Pro, $29 a monthVerbatim quotes a short extract and links to the episode; the episode belongs to its publisher. Remove a show.