Transforming Search with Perplexity AI’s CTO Denis Yarats
Gradient Dissent: Conversations on AI
00:00
Model Selection Preferences and Model Development at Perplexity AI
The chapter explores user preferences for diverse model selections in search engines, discussing the future automated selection of models based on query complexity and compute power. It covers the development of post-trained open source models at Perplexity AI for generating search results and answers, focusing on optimizing detailed and nuanced responses. The conversation also touches on the process of refining training data quality, establishing guidelines for good answers, and addressing challenges in ongoing training and AI model retraining.
Transcript
Play full episode
