19 May 2021
Google on the first day of its digitally held I/O 2021 event announced LaMDA, its latest research in the language model of machine learning.
LaMDA is Google’s latest effort in making machine learning feel more conversational and natural than before. Unlike most other language models, LaMDA aims at establishing an open-ended conversation that could feel more natural and chatty. Google puts up an example that how a chat with a friend about a TV show sometimes dwells further and they end up talking about something completely different from where they started, like what was the country where the show was filmed before.
An open-ended can start at one topic and evolve further and further. On the stage, Google demoed how AI actively participated in the conversation about the planet Pluto by pretending to be Pluto.
Every new language model advances the way AI interacts with us. The aim has always been to make it sound more humanly as possible and LaMDA is a “breakthrough” in this. Of course, the conversation is not going to be as colloquial as what we humans do but it is definitely better.
Google claims LaMDA can converse without any kind of prior training.
It is definitely time that digital assistants should serve better, after all, no one wants it to be just limited to setting up alarms, reminders, or play questionaire.