3 ms·
>they don't have to add anything that wasn't already expressed in the paper, limiting the need for creativity But how could we know they add nothing or missed
by croes 3y ago
>they don't have to add anything that wasn't already expressed in the paper, limiting the need for creativity
But how could we know they add nothing or missed something important like a not?
>they can give a more neutral overview of the research since they aren't an involved entity with emotional stake
Biased input = biased output.
AI can't help with that.
- squigz 3y ago> But how could we know they add nothing or missed something important like a not? Probably by reading it.
- croes 3y agoSo I read the original after the abstract? Sounds like more work not less.
- yu3zhou4 3y agoHere's how I picture this, in an optimistic scenario: Once the number of abstracts grows up, we vote up and down on them, so at the end you get the best abstract for each paper, which is supposed to be informative enough to give you a gist of what authors have learned from their research. It should be enough to let you keep an eye on what's going on in research world in real time and also serve as a filter on what paper is worth reading next This is a grand vision of course, right now we have 9 abstracts, but hey, it's still a good starting point! :p
- squigz 3y agoNo... the author reads it before publishing...
- wongarsu 3y agoPresumably you read the summaries to find promising papers where you will read the abstract (which you then follow up by reading the figures, conclusion, discussion, introduction, then the rest, bailing out once you think this isn't a paper you want to read further). If at any point you figure out the "community abstract" was misleading you can come back and downvote it. The only issue is if community abstracts get published that make good papers unappealing to read, or if the majority of votes come from people who don't actually engage with the paper
- diffeomorphism 3y agoBetter than human is good enough. Also, you will be limited by the texts themselves, which can oversell or just make wrong claims. The main advantage I am looking forward to is scaling. As a human there is only so much complex text I can read in 24h. If I can just ask my computer to skim 200 articles and tell me which ones seem worth really digging into that would be awesome.
- yu3zhou4 3y agoYes! After finishing the most urgent features like sorting, filtering and fixing some bugs, I'd like to make it easy to subscribe to a chosen subject to get a relevant papers to your mailbox