Last Week in AI The Alexa Prize Challenge with Stanford's Abigail See and Ashwin Paranjape
Aug 20, 2020
Abigail See and Ashwin Paranjape, PhD students at Stanford AI Lab, share insights on their award-winning chatbot, Chirpy Cardinal, which secured second place in the Alexa Prize Challenge. They discuss the importance of rapid development and user feedback in creating engaging conversational AI. The duo explores the mixed initiative approach used in Chirpy Cardinal, enhancing its ability to generate empathetic responses while navigating complex social topics. They also touch on future innovations in AI and the role of open-source resources in advancing conversational interfaces.
AI Snips
Chapters
Transcript
Episode notes
User Testing
- Users tested Chirpy Cardinal through Alexa devices.
- Saying "Let's chat" routed users to competing chatbots.
User-Driven Conversations
- The chatbot prioritized user initiative, letting users drive conversations.
- This approach is more engaging but harder to implement effectively.
Topic Switching
- Chirpy Cardinal could switch topics based on entities users mentioned.
- Mentioning "Doctor Who" could trigger a conversation about the show, using Wikipedia data.