Lloyd Frank
A
I've been playing around a bit with ChatGPT as a way to test some possibilities for class discussions. I'm hung up, though, on the ethical problem of ChatGPT's training dataset comprising, in part, copyrighted material used without permission. Thoughts?
10:35 AM - Apr 01, 2023
Avatar
0
3
1
#Glassbydeanna
A
In response to Lloyd Frank.
Copyright issues need to be addressed now or it will cause compounded issues as time goes on. Using the web as a data set gives a highly skewed “normal” that I do not see anyone talking about. Anything published historically was generally written by wealthy males.Same with most info on the web now.
10:43 AM - Apr 01, 2023
1
1
Lloyd Frank
A
In response to #Glassbydeanna.
Totally agree. For now, OpenAI is trying to hide behind the "fair use" clause, but that only covers 10% or 1000 words of a work. Not ingesting the whole thing. Also, unless the knowledge is in digital form, it is not part of the dataset. That seems particularly limiting.
10:58 AM - Apr 01, 2023
0
1
RA
A
In response to Lloyd Frank.
Yep, that's my issue too, and it's impossible to recognize where it all comes from, even though some experts would be able to tell. Also, who owns the output? I bet it's not the public.
10:39 AM - Apr 01, 2023
0
1

 

{{ notificationModalContent }} {{ promptModalMessage }}