Google Assistant has become synonymous with natural language understanding, and Google’s latest technological breakthrough “LaMDA” — a language model for natural conversation apps — will bring the future even more within reach.

Following “BERT “, Google announced “LaMDA “today at I/O 2021 as a breakthrough in development, with a focus on: natural dialogue.

LaMDA is our latest research breakthrough, adding pieces to one of the most tantalizing pieces of the puzzle: conversation. While conversations tend to revolve around specific topics, their open-ended nature means they can start in one place and end up in completely different places. A conversation with a friend about a TV show can turn into a discussion of the country where the show is filmed, which then leads to a discussion of the best regional cuisine in that country. This meandering nature quickly bogged down modern conversation agents (commonly known as chatbots), which tend to follow narrow, pre-defined paths. But LaMDA– short for “Language Model for Conversational Applications” — has the ability to discuss seemingly endless topics in a free-flowing way, and we think this ability could unleash more natural ways of interacting with technology and a whole new class of beneficial applications.

In a brief demo at I/O 2021, Google showed LaMDA in action as Pluto and paper airplane. In each case, LaMDA has a strong understanding of both topics and can respond as the object when asked a question. For example, when asked “What else do you want people to know about you”, LaMDA’s Pluto can answer: “I want people to know that I’m more than just a random puck. I’m actually a beautiful planet.”

To that end, Google says it trains the technology in conversations.

But unlike most other language models, LaMDA is trained in conversation. During his training, he discovered some nuances that distinguish him from other forms of open conversation. One such nuance is sensitivity. That’s basically it. Does the response to a given conversational context make sense?

LaMDA isn’t available in any products yet, but Google is working on the technology that could eventually be used in products like assistant, search and Workspace. In a blog post, Google noted that it was working to ensure LaMDA’s answers were both “compelling” and “correct.”

The original link: 9 to5google.com/2021/05/18/…