Bernhardsson Links LLM Limits to Short Feedback Loops
Erik Bernhardsson shares idea on why intelligent people and models may underperform.
Erik Bernhardsson posted a theory that highly intelligent people sometimes underperform because they focus on verifiable rewards with short feedback loops. He suggested the same pattern could explain limits in large language models. Researcher Shreya Shankar shared the post, commenting that it illustrates how grad students and academics engage with ideas on social media. The packet contains only these two posts and the generated headlines summarizing them.
I've had a theory about super smart people who aren't always successful, that it boils down to focusing too much on verifiable rewards with short feedback loops. It feels like there's a lesson that might carry over to LLMs here.
Combined views
Reactions
4 postsSS 
Shreya Shankar@sh_reyaGrad students and academics 🤝 social media
- likes: 95
- replies: 4
- bookmarks: 43
- reposts: 2
CY Chhavi Yadav@chhaviyadav_Being patient, keeping faith and ability to be sane under uncertainty kids
- likes: 6
- replies: 0
- bookmarks: 1
- reposts: 0
SC Siqi Chen@bladerthere are so many powerful lessons to learn when you think about your own mind as an LLM
- likes: 28
- replies: 4
- bookmarks: 8
- reposts: 1
EB Erik Bernhardsson@bernhardssonI've had a theory about super smart people who aren't always successful, that it boils down to focusing too much on verifiable rewards with short feedback loops. It feels like there's a lesson that might carry over to LLMs here.
- likes: 2.5K
- replies: 126
- bookmarks: 1.3K
- reposts: 99