Google Unveils LaMDA and MUM AI Models to Revolutionize Conversation and Search
Google introduces LaMDA, an open-domain language model built for natural dialogue, alongside MUM, a multimodal system designed to understand text and images across 75 languages to answer complex questions.
Google reveals LaMDA, a new language model built specifically for dialogue applications that surpasses current models in understanding conversational context. Unlike traditional systems, LaMDA operates as an open-domain model that converses on any topic without requiring retraining or predefined responses, generating natural and unique interactions every time.
In addition to LaMDA, Google introduces the multitask unified model (MUM), a multimodal AI that processes text, images, and videos simultaneously across 75 languages. This allows MUM to handle complex, nuanced questions by drawing connections between different types of information, such as combining fitness advice with weather recommendations for planning a hiking trip.
Although Google admits LaMDA occasionally produces nonsensical responses during this early research phase, the company actively tests both AI models through internal pilots. Together, these technologies represent a major step forward in Google's ability to deliver more intuitive and context-aware search experiences that understand both the intent and the visual elements of user queries.