this post was submitted on 03 Jan 2024
262 points (100.0% liked)

196

667 readers
71 users here now

Be sure to follow the rule before you head out.

Rule: You must post before you leave.

^other^ ^rules^

founded 1 year ago
MODERATORS
moss@lemmy.blahaj.zone
greembow@lemmy.blahaj.zone
queue
funky_rodent@lemmy.blahaj.zone
PeachyMcPeachface@lemmy.blahaj.zone
remotelove@lemmy.ca
Roflmasterbigpimp@feddit.de
A_Very_Big_Fan@lemm.ee
qaz@lemmy.blahaj.zone
qaz@lemmy.ca
qaz@lemmy.dbzer0.com
Qaz@lemmy.ml
qaz@lemmy.sdf.org
262
AI grammar rule (lemmy.blahaj.zone)
submitted 10 months ago by FakeGreekGirl@lemmy.blahaj.zone to c/196@lemmy.blahaj.zone
11 comments fedilink hide all child comments
 

https://bsky.app/profile/kjhealy.bsky.social/post/3ki3kq6bhad2u

you are viewing a single comment's thread
view the rest of the comments
[โ€“] peanuts4life 35 points 10 months ago (1 children)

I've seen this with gpt4. If I ask it to proofread text with errors it consistently does a great job, but if I prompt it to proofread a text without errors, it hallucinates them. It's funny to see Microsoft having the same issue.

[โ€“] FakeGreekGirl@lemmy.blahaj.zone 25 points 10 months ago

I'm pretty sure MS uses GPT-4 as the foundation of all their AI stuff, so it's not surprising to see them have the same issues. Funny, as you said, but not surprising.