From using AI to predict what users want and help them complete tasks, Google is now taking steps forward to make the AI friendly. And for a machine to be your friend, it must be able to hold conversations by speaking nicely. Until now, Google assistant It used to sound like a robot voice just announcing search results. But here comes LaMDA or ‘Language model for dialogue applications‘- to make Google new “breakthrough” technology Android Phones, smart speakers, cars, and other voice-enabled devices can talk like humans.
With LaMDA, the Google Assistant can have engaging conversations with you, “in a free-flowing manner, on a seemingly endless number of topics”. So when you ask the Google Assistant about the current weather, it just keeps saying “35 degrees Celsius, sunny and 65% humidity”.
Google will respond more kindly to everyday questions like these, such as: B. “It’s warm outside, you might want to take a bottle of water with you”. Interestingly, it doesn’t stop there. Google will try to make it more attractive by continuing to talk to you. For example, the conversation could turn into “Hey, it can get a little cooler in the evening, maybe you want to enjoy a beer in your favorite place.” And from the beer talk, you might move on to movies, food, or any other topic that you’re interested in. This is similar to the way people talk to each other.
Google mentioned in an official blog post that the whole idea of LaMDA mimics human conversation. “Talking to a friend about a TV show could turn into a discussion about the country the show was filmed in before debating about the best local food in that country,” it said.
What is the technology behind LaMDA?
LaMDA’s “Conversational Skills” are a long research project and will continue to evolve. “Like many newer language models, including BERT and GPT-3 is based on Transformer, a neural network architecture that Google Research invented and open source in 2017, “said Google.
The older architecture creates a model that can be trained to read many words (e.g., a sentence or paragraph), pay attention to how those words relate to one another, and then predict which words will come next become.
But LaMDA is different. It is trained in dialogue. “During training, it picked up on some of the nuances that distinguish open conversation from other forms of speech,” it added. Whatever the AI is talking has to make sense and whenever you are talking to a LaMDA-assisted voice AI the overall goal is to think before you respond.
“Basically: Does the answer make sense to a specific conversation context?”
Google continues to try to get the LaMDA architecture to respond to people with interesting answers. “We also examine dimensions such as ‘interesting’ by assessing whether the answers are insightful, unexpected, or funny. As Google, we place great value on facts (ie, whether LaMDA sticks to facts that language models often struggle with) and are exploring ways to ensure that LaMDA’s answers are not only convincing but correct, ”said Google.