this post was submitted on 06 Jul 2023
671 points (94.3% liked)
ChatGPT
8901 readers
1 users here now
Unofficial ChatGPT community to discuss anything ChatGPT
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 think the ethical filters and other such moderation controls are hard coded pre-process thing. That's why it repeats the same things over and over, and has the same hangs up as early 00s poorly made censor lists. It simply cuts off the system and substitutes a cookie cutter response.
I find it interesting that they don't offer a version of GPT 4 that uses it's own language processing to screen responses for "unsafe" material.
It would use way more processing than the simple system you outlined above, but for paying customers that would hardly be an issue.
It's possible but also incredibly complicated and technically involved to tweak a LLM like that. It's one of the main topics of machine learning research.
A lot of people get stuck with issues like that where there are conflicting principles.