I've become a superpredictor just through this rule of thumb: future AI will improve faster than you think, but current capabilities are always exaggerated. Another example from @danishpruthi.bsky.social

Yep! To add to the insult, we know that a lot of "AI-generated original research ideas" are plagiarised. I ended up writing a blog post about it jilltxt.net/can-an-ai-ag... - see also @danishpruthi.bsky.social thread on X about their study finding the plagiarism x.com/danish037/st...

My #ICLR2025 story: Reviewed 3 papers. Blocked a couple of hours to engage with authors. Of the three papers, 1 is withdrawn, and 2 have no author responses (for any of the reviews). And yes, no reviewer engagement on the paper we submitted.