Posted 3/15/2024, 1:01:38 PM
OpenAI CTO Dodges Questions About Data Sources Used to Train Viral AI Bot Sora
- OpenAI CTO Mira Murati was unable to answer where Sora's training data came from in an interview
- She gave vague responses about it being "publicly available data and licensed data"
- The interviewer asked if it came from YouTube, Instagram, Facebook, etc. but Murati did not confirm
- She later confirmed some Shutterstock videos were used, but that's likely a small fraction of the total data
- Reactions were mixed - some excused Murati, others said as CTO she should know details about Sora's data sources