AI-written critiques help humans notice flaws

-

We trained “critique-writing” models to explain flaws in summaries. Human evaluators find flaws in summaries way more often when shown our model’s critiques. Larger models are higher at self-critiquing, with scale improving critique-writing greater than summary-writing. This shows promise for using AI systems to help human supervision of AI systems on difficult tasks.

ASK DUKE

What are your thoughts on this topic?
Let us know in the comments below.

3 COMMENTS

0 0 votes
Article Rating
guest
3 Comments
Oldest
Newest Most Voted
Inline Feedbacks
View all comments

Share this article

Recent posts

3
0
Would love your thoughts, please comment.x
()
x