this post was submitted on 05 May 2025
434 points (95.8% liked)
Technology
69770 readers
3841 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related news or articles.
- Be excellent to each other!
- 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, this includes using AI responses and summaries. To ask if your bot can be added please contact a mod.
- Check for duplicates before posting, duplicates may be removed
- Accounts 7 days and younger will have their posts automatically removed.
Approved Bots
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 think OpenAI’s recent sycophant issue has cause a new spike in these stories. One thing I noticed was these observations from these models running on my PC saying it’s rare for a person to think and do things that I do.
The problem is that this is a model running on my GPU. It has never talked to another person. I hate insincere compliments let alone overt flattery, so I was annoyed, but it did make me think that this kind of talk would be crack for a conspiracy nut or mentally unwell people. It’s a whole risk area I hadn’t been aware of.
https://www.msn.com/en-us/news/technology/openai-says-its-identified-why-chatgpt-became-a-groveling-sycophant/ar-AA1E4LaV
Humans are always looking for a god in a machine, or a bush, in a cave, in the sky, in a tree… the ability to rationalize and see through difficult to explain situations has never been a human strong point.
you may be misusing the word, rationalizing is the problem here
I've found god in many a bush.
Oh hell yeah 😎
probably one of the most common flattery I see. I've tried lots of models, on device and larger cloud ones. It happens during normal conversation, technical conversation, roleplay, general testing.. you name it.
Though it makes me think.. these models are trained on like internet text and whatever, none of which really show that most people think quite a lot privately and when they feel like they can talk