We're sunsetting PodQuest on 2025-07-28. Thank you for your support!
Export Podcast Subscriptions
Dr. Margaret Mitchell shares her experiences in AI research, revealing how AI systems made unsettling errors, such as describing a disaster scene as 'awesome.' These errors highlight the importance of understanding the data these systems are trained on.
- Dr. Margaret Mitchell worked on vision-to-language AI, teaching systems to describe images.
- The systems made errors, like calling an explosion scene 'awesome,' due to biased training data.
- These AI errors sparked concerns about the potential consequences if such systems were widely deployed.
- Dr. Mitchell realized that the AI was a reflection of the biases present in the training data.
Shownotes Transcript