Michael Ten @lemmy.world to Technology@lemmy.worldEnglish · 9 months agoOpenAI transcribed over a million hours of YouTube videos to train GPT-4www.theverge.comexternal-linkmessage-square41fedilinkarrow-up199arrow-down19cross-posted to: technology@lemmy.ml
arrow-up190arrow-down1external-linkOpenAI transcribed over a million hours of YouTube videos to train GPT-4www.theverge.comMichael Ten @lemmy.world to Technology@lemmy.worldEnglish · 9 months agomessage-square41fedilinkcross-posted to: technology@lemmy.ml
minus-squareLvxferre@mander.xyzlinkfedilinkEnglisharrow-up22arrow-down1·9 months agoThe “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.
minus-squareGrimy@lemmy.worldlinkfedilinkEnglisharrow-up2arrow-down1·9 months agoIf 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.
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.