Writing is fundamentally the transfer of information from your brain to my brain. If you have 1000 bits of semantic information you want to transfer, you can't give 300 bits of semantic information to an LLM and have it fill in the remaining 700, because it doesn't know what those 700 bits are. If it's able to guess those 700 bits correctly, then they aren't true semantic information, and you really only have 300 bits you want to transfer. You might as well transfer those bits to me directly, rather than having the LLM add on an extra superfluous 700 bits that I then have to filter out.
Saying something for years doesn't make it right. Has nobody ever challenged you on that? Our brains prefer to read enjoyable bits, not raw information. If you only got those 300 bits, you aren't going to bother putting them into your LLM to generate 700 bits to make it a more enjoyable read, you're just going to struggle through the 300. Maybe when browsers come with built-in automatic text puff-uppers, then you'll have a case but almost nobody does that now.
Here's a clearer example - would you rather learn a concept from a research paper or a textbook or blog? You say the research paper but they're dense and hard to wade through where-as blogs and textbooks are more wordy but hold your hand, which is something that helps humans learn.
Plenty of people have challenged me on this! There are a lot of exceptions we can come up with (summarization and private context are the biggest examples), and everything's changing in various ways as AIs get more capable. Usually, conversations about people pasting LLM output focus on things like AI generated blog posts, people who have no taste in writing and want an AI to rewrite everything for them, people who are just plain lazy, etc.
> would you rather learn a concept from a research paper or a textbook or blog?
I pretty much always read blogs first, and then move to a research paper only if I want more details or care enough about the subject to verify with the original source. Typically this is because research papers have too much information to be approachable.
hatthew · · focus · HN ↗
Writing is fundamentally the transfer of information from your brain to my brain. If you have 1000 bits of semantic information you want to transfer, you can't give 300 bits of semantic information to an LLM and have it fill in the remaining 700, because it doesn't know what those 700 bits are. If it's able to guess those 700 bits correctly, then they aren't true semantic information, and you really only have 300 bits you want to transfer. You might as well transfer those bits to me directly, rather than having the LLM add on an extra superfluous 700 bits that I then have to filter out.
foxglacier · · focus · HN ↗
Here's a clearer example - would you rather learn a concept from a research paper or a textbook or blog? You say the research paper but they're dense and hard to wade through where-as blogs and textbooks are more wordy but hold your hand, which is something that helps humans learn.
hatthew · · focus · HN ↗
> would you rather learn a concept from a research paper or a textbook or blog?
I pretty much always read blogs first, and then move to a research paper only if I want more details or care enough about the subject to verify with the original source. Typically this is because research papers have too much information to be approachable.