this post was submitted on 21 Jul 2023
256 points (100.0% liked)
Technology
37708 readers
506 users here now
A nice place to discuss rumors, happenings, innovations, and challenges in the technology sphere. We also welcome discussions on the intersections of technology and society. If it’s technological news or discussion of technology, it probably belongs here.
Remember the overriding ethos on Beehaw: Be(e) Nice. Each user you encounter here is a person, and should be treated with kindness (even if they’re wrong, or use a Linux distro you don’t like). Personal attacks will not be tolerated.
Subcommunities on Beehaw:
This community's icon was made by Aaron Schneider, under the CC-BY-NC-SA 4.0 license.
founded 2 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
I was curious how current LLMs might handle this with proper instructions, so I asked chatGPT this: “What can you tell me about this Reddit post? Would you write a news article about this? Analyze the trustworthiness of this information:” and pasted the text from the post. Here’s a part of its reply:
So it’s not even an issue with current models, just bad setup. An autoGPT with several fact-checking questions added in can easily filter this stuff.
That’s not the right approach.
CharGPT is going off of old information, something not good for “breaking” news.
An example is not knowing Dregonflight is the name of the WoW expansion from late 2022.
So theoretically Glorbo could be a real character but only known more recently or that stout guy could’ve had a character arc that makes it a possible outcome yet current LLMs failed to pick up on that.
Half of the deleted […] things are chatGPT mentioning its 2021 knowledge cutoff and suggesting double-checking that info. It was mentioned in this case as well.
If it were an autoGPT with internet access, I think these would prompt an automated online lookup to fact-check it.
Try these questions about sources of recent information that you believe are accurate.
So I tried it on this BBC article (a current top story), and this /r/Hearthstone post. It did pretty well. I won't copy-paste the whole reply, but here are some excerpts:
So it guessed correctly in both cases and suggested where to fact-check the info to be sure.
Did you intend to paste or attach something? Your comment doesn't show anything on kbin besides that one sentence.
I think you may be misunderstanding this comment (no shade). I think they’re not saying “try these (that I will now provide) questions”. They’re saying “try these questions (that you asked in your previous query), and ask those same questions about sourced material that you trust or believe to be true.”