Dwarkesh is on fire atm! Sample efficiency, continual learning, OPSD, World Models, etc. But data privacy will be a blocker. We need personalized models! https://www.dwarkesh.com/p/the-next-paradigm
5:59 PM · Jun 26, 2026 · 14K Views
Dwarkesh Patel argues the next major AI breakthrough will be systems that learn continually on the job
Kevin Patrick Murphy warns privacy constraints will block this.
811386215.1K
Sentiment
Positive users praised the Dwarkesh Podcast episode on sample efficiency, world models, and personalized AI for its insightful content that motivates further learning.
Pos
100.0%
Neg
0.0%
2 comments with sentiment.
Cluster Engagement
Digg Deeper
No Digg Deeper questions have been answered for this story yet.
Related links
Posts from X
Most Activity
Most Activity
VIEWS1.4KLIKES14REPLIES1
Dwarkesh Patel@dwarkesh_sp
@sirbayes Glad you liked, must mean I didn't get it all wrong ha
Kevin Patrick Murphy@sirbayes
Dwarkesh is on fire atm! Sample efficiency, continual learning, OPSD, World Models, etc. But data privacy will be a blocker. We need personalized models! https://www.dwarkesh.com/p/the-next-paradigm
8hViews 1.4KLikes 14Bookmarks 0
T Shirt n Jeans@TShirtnJeans2
@sirbayes Personalized models? If memory wasn't in a shortage already
8hViews 23
ツ Interesting Name ツ@Interestingna11
@dwarkesh_sp @sirbayes This is so good, I’m going on a bender tonight to learn a lot more.
Thank you, Indian Gigachad 💪
8hViews 3
Taro Bushidō@techietaro
@sirbayes RL's sparsity as a feature for continual learning is a sharp insight. But OPSD assumes the teacher model's in-context learning holds at long horizons. Dario's comment suggests it might not. What then?
8hViews 2