Google’s AI website:  https://ai.google/

Google’s Vision:

Language is a complex and adaptable tool that has always been of interest to technology companies like Google. In the past, Google has focused on translating the web and developing machine learning techniques to understand search queries. While these advances have made it easier to access and organize written and spoken information, there is always room for improvement.

One of the most tantalizing puzzles in computer science is conversation, which is one of the most nuanced and versatile forms of language. Conversations can start in one place and end up somewhere entirely different, which can make it difficult for modern conversational agents or chatbots to follow narrow, predefined paths. However, Google’s latest research breakthrough, LaMDA, short for “Language Model for Dialogue Applications,” can engage in a free-flowing way about an endless number of topics. This ability could unlock more natural ways of interacting with technology and new categories of helpful applications.

LaMDA is built on the Transformer neural network architecture that Google Research invented and open-sourced in 2017, which produces a model that can read many words, pay attention to how those words relate to one another, and predict what words it thinks will come next. However, unlike most other language models, LaMDA was trained on dialogue, picking up on the nuances that distinguish open-ended conversation from other forms of language.

LaMDA’s training includes sensibleness, which is whether the response to a given conversational context makes sense, and specificity, which relates clearly to the context of the conversation. LaMDA also focuses on factuality and interestingness, ensuring that its responses are insightful, unexpected, witty, and correct.

While LaMDA’s conversational skills are impressive, Google’s highest priority is ensuring that the technology adheres to AI principles and does not propagate biases, hate speech, or misleading information. To minimize these risks, Google builds and open-sources resources for researchers to analyze models and the data on which they’re trained, scrutinizes LaMDA’s development, and continues to work on incorporating conversational abilities into more products.