Google Unveils LaMDA and MUM to Advance Conversational AI
Google introduces LaMDA and MUM at I/O 2021 to make AI conversations feel more human and flexible. These new language models aim to understand context and handle open-ended dialogue without needing retraining.
Google unveils two new major language models, LaMDA and MUM, at its virtual I/O 2021 event to significantly improve conversational AI. Although Google Assistant receives little direct attention during the presentation, these new tools show how the company plans to make AI interactions feel much more like talking to a real person.
LaMDA, or Language Model for Dialogue Applications, focuses specifically on handling open-ended conversations that naturally drift from one topic to another. Built upon the foundation of the earlier Meena chatbot, LaMDA analyzes entire paragraphs and sentences to grasp the broader context, allowing it to switch seamlessly from discussing Pluto to explaining how to make a paper airplane without any retraining.
MUM serves as a multitask unified model designed to make AI more helpful when users rely on casual phrasing or complex queries. Despite the impressive capabilities of these open-domain models, some experts express skepticism about whether LaMDA maintains sensible and satisfying responses across every possible subject during long-term use.