OpenAI transcribed over a million hours of YouTube videos to train GPT-4
OpenAI transcribed over a million hours of YouTube videos to train GPT-4
www.theverge.com OpenAI transcribed over a million hours of YouTube videos to train GPT-4
How OpenAI, Google, and Meta deal with the limits of data online.
![[News] OpenAI transcribed over a million hours of YouTube videos to train GPT-4](https://lemmy.world/pictrs/image/23969ad4-7554-4cef-b62e-68b326a178b2.jpeg?format=webp&thumbnail=196)
You're viewing a single thread.
43 comments
The "fun" part is that they'll get away with it. Google likely did the same with other platforms; so if it sues OpenAI, it's creating a big precedent against itself.
If they didn't, google would own all video data just like Reddit would own most conversation data.
It's a good thing, at least this lets the open source scene be a player.
In what world is OpenAI open source?
They aren't but if they can disregard ToS, so can the open source scene.
43 comments