Post

Craig Reynolds
@craigreynolds.bsky.social

ACM Digital Library added an “LLM first” mode like this in a recent redesign. Not my cup of tea, but I don't object in principle to providing it as opt-in. They screwed up badly at first, setting “AI mode” to be ON by default, which lead to many very loud complaints. It has since been fixed.

Richard Sever
@richardsever.bsky.social

One thing you give to potential editors in hiring tests is a badly written paper whose abstract is barely comprehensible and omits key info. Generally you don’t have to look far to find one… 1/2

Richard Sever
@richardsever.bsky.social

I’m hardly an AI lobbyist but we have to be honest LLMs will write summaries better than many authors, especially for lay readers. Feels like one thing they’re inarguably good at (though not perfect given hallucinations etc).

Jeremy Yoder says no war
@jbyoder.org

So (A) even if I agree that LLM summarization is sufficiently more reliable than all authors everywhere, just generate the summary and drop it in front of the paper, don’t waste the compute and my time with a chatbot; and

Jeremy Yoder says no war
@jbyoder.org

There is no market for a paper-reading chatbot embedded with every scholarly article, much less a paper-reading chatbot that needs a disclaimer about how it does not in fact always read the paper correctly