On-Device Data Solutions
Joe discusses the potential of on-device models for enhancing privacy by keeping data local, allowing for functions like summarizing messages without cloud dependency. He also addresses the challenges of data availability, suggesting that synthetic data could be a viable solution moving forward. While acknowledging the advantages larger companies have in data access, he remains optimistic about ongoing improvements in model performance and data generation.In this clip
From this podcast
Training Data
Meta’s Joe Spisak on Llama 3.1 405B and the Democratization of Frontier Models | Training Data
Related Questions