Latent Space: The AI Engineer PodcastLatent.SpaceAll episodes
We first had Nathan on to give us his RLHF deep dive when he was joining AI2, and now he’s back to help us catch up on the evolution to RLVR (Reinforcement Learning with Verifiable Rewards), first proposed in his Tulu 3 paper. While RLHF remains founda...
Full show notes and links
A podcast does not come with subtitles, so podqast makes them: it listens to the episode itself, a few minutes at a time, and only around where you are listening. Nothing is done in advance, which is why the bar says how much of the episode you can read so far.
Each stretch is heard by a speech model, cut into lines at the pauses and full stops, and translated in one go — what was said on top, English under it, and the two lit together piece by piece as they are spoken.
It is a machine doing it, so it is wrong sometimes. What was said is always the line on top.