this post was submitted on 20 Mar 2024
53 points (88.4% liked)
Technology
59589 readers
2838 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related content.
- Be excellent to each another!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, to ask if your bot can be added please contact us.
- Check for duplicates before posting, duplicates may be removed
Approved Bots
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
I can't trust the output of an LLM but at least you can ask it to cite its sources so you can get the page that helped it come to that conclusion
If I ask an LLM something like “is there a git project that does <something I’d describe in natural language but not keywords>” or is there a Windows program that does X, it may make up the answers, but at least I can verify that via a search engine. If I try to Google that (that’s theoretical, I don’t use Google), I’m going to end up on a page full of ads that is filled with trap links that lead to malware, top 10 lists with the same repeated content and all the other shit the internet has become. I kind of don’t mind the hallucinations relative to the ads. What a time to be alive though.
Obviously it depends on the LLM, but ChatGPT Plus doesn't hallucinate with your example. What it does is provide a list of git projects / windows programs, each with a short summary and a link to the official website.
And the summary doesn't come from the website — the summary is a short description of how it matches your requirements list.
I've also noticed Bing has started showing LLM summaries for search results. For example I've typed a question into Duck Duck Go (which uses Bing internally) and seen links to reddit where the answer is "a user answered your question stating X, and another user disagreed saying Y".
I'm encountering hallucinations far less often now than I used to - at least with OpenAI based products.
That’s a really good use case that I will need to start using. In a use case such as that where paid ads make the search engines unreliable, the LLM is at least going to be on the same footing if not better.
You immediately start from a better place using an LLM vs searching raw. If I just need a link, I use DDG. When I need to research, I ask the latest gpt4-turbo model.