Unsupervised Learning with Jacob Effron · Ep 94: Applied Compute CEO on the Limits of RL, the New AI Hyperscaler & Why Post-Training Wins Inference
Audio streams from the publisher’s own feed. Verbatim does not host or copy it.
adopt an inverted-pyramid strategy: train at the bottom, then inference, routing, context, harness
“Jev is a router… as I have text, route to the right tool, and then if I can infer from the text, the job to do in the tool, I might as well infer that and route as far down the user journey as I can.”
“The cost to post-train ... is falling. You're still better off doing a bunch of context optimization and harness optimization, but once you squeeze a lot out of the harness you can continue to squeeze a lot out of optimizing the policy”
“Seed RL… was about utilizing the TPUs both for inference and training in the best possible way while keeping the environments as flexible as you wanted.”
“… you can prompt it with an image… or with video… you can do it in context… you don’t basically need to train it.”
The strongest new moments on your topics, in the speakers' own words. Free for three topics.
No card. No password. One-click unsubscribe.
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.