All the LLMs would like to apologize for not writing a long response, they...
All the LLMs would like to apologize for not writing a long response, they didn't have time to write a short one.
Semantic, lossy compression is hard. There are few non-AI ways to do it
- gensim summarization
- replacing phrases with acronyms
- stripping out decorative formatting (e.g. html tags)
- text message speak (may make answers worse!)
- ... uh.... running out of ideas
Self-replies
It sucks that the leading option for summarization is to give a cheap bot a text to summarize, possibly repeatedly because they estimate poorly how much they've compressed, then give the summary to a more expensive bot.
Maybe "skimming"? Humans will take a random sample of sentences to get a summary without reading everything.