this post was submitted on 08 Jun 2024
231 points (96.4% liked)
Movies and TV Shows
2114 readers
41 users here now
A community for entertainment industry news and general discussion about movies and TV shows.
Rules:
- Be civil.
- Please do not link to pirated content.
- No spoilers in the title of submissions. And please use spoiler MarkDown in the body of discussions. This is a courtesy to other users.
- Comments solely criticizing headlines and/or journalism will be removed for being off-topic.
founded 1 year ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
Sometimes just cutting out paragraphs to "summarize" the text is just plain bad. This is one of those times, and time for me to finally block this thing.
But but suddenly critics it's but. I'm not defending the bot this is just the first words of each paragraph and it's really annoying to read.
You realise you are reacting with feelings to a bot? I read the whole article and it's just an interview format which makes no sense to summarise. I mean you can always disregard, but it does save some time most of the time, helping you quickly judge if the content is worth a full read or not without opening the link. It's a technology
I am very aware of that it is a bot 😄. And I have to say: a rather bad-its-job one, not just in this case.
Whenever I read one of its "summaries" it was blatantly obvious that it just cuts parts, often times mauling paragraphs and destroying contexts. And that does not offer any value to me, hence I permanently disregard it now by putting that account on a block list (next to other not-so-great bot accounts).
that’s actually what the underlying method does, as this is extractive summary, hence it mostly cuts and stitches things.
From my naive understanding, this type of method does not use or “understand” context.
The alternative is abstractive summary, which is where LLMs (or even small/medium language models) are good for. But I suspect that would be a controversial choice on lemmy.
I think that this is the logical place to put those opinions - it would make less sense unlinked to the post.
I dislike that it's been shotgunned at nearly every post I see, regardless of how relevant it might be, and I don't trust the output to represent the link.
Given there's no "bad bot" style feedback functionality (that I know of), what else is there to do than block it?
Github feedback? The bot's code didn't write itself. I doubt the programmer(s) read all the posts that the bot generates