this post was submitted on 03 May 2024
25 points (62.9% liked)
Asklemmy
43939 readers
859 users here now
A loosely moderated place to ask open-ended questions
Search asklemmy ๐
If your post meets the following criteria, it's welcome here!
- Open-ended question
- Not offensive: at this point, we do not have the bandwidth to moderate overtly political discussions. Assume best intent and be excellent to each other.
- Not regarding using or support for Lemmy: context, see the list of support communities and tools for finding communities below
- Not ad nauseam inducing: please make sure it is a question that would be new to most members
- An actual topic of discussion
Looking for support?
Looking for a community?
- Lemmyverse: community search
- sub.rehab: maps old subreddits to fediverse options, marks official as such
- !lemmy411@lemmy.ca: a community for finding communities
~Icon~ ~by~ ~@Double_A@discuss.tchncs.de~
founded 5 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
That is a very popular, but also very dangerous misconception. AI has all the same biases and opinions as the dataset it was trained on (and thus also those of the engineer who picked said dataset). Even if you just YOLO it by training it on "everything" and hoping it'll average out, whatever biases society itself as a whole has, the AI will happily perpetuate.
For example, folks wanted to reduce judge bias in criminal sentencing, so they created an AI... and then trained it on historical sentences. Guess what happened?
In reality, if you want to create an unbiased AI, you've got to go out of your way to carefully curate the dataset to deliberately remove bias, and almost nobody is doing that.