|
I’m ok with AI generating all future tv and movies from 90s scripts.
|
# ¿ Feb 1, 2023 18:01 |
|
|
# ¿ Mar 28, 2024 11:28 |
|
I'm not sure what language model Nothing Forever was using but the base models are basically just predicting the next word to form sentences from massive amounts of text data on the internet/social media. The data itself can have offensive content or there can be topics that are rare in the text data that the AI doesn't know how talk about when it comes up. As I understand it, ChatGPT tweaks this base model with human reinforcement where a group of people will tell the AI how good/bad its responses were and the AI adjusts the model. And ofc you can have filters in the text training data as well as the AI output. OpenAI even has their own moderation tools that check the output for hateful or harmful content. I have no idea how much of this is free though for someone just doing an AI project like Nothing Forever.
|
# ¿ Feb 7, 2023 20:51 |