New AI technology should make Google Assistant more human

Google Assistant will soon talk more naturally, and other than just summing up dry facts, Google can have a real conversation. The new ‘LaMDA’ technology based on artificial intelligence (AI) must get the Assistant working.

‘LaMDA’ artificial intelligence

If you ask the Assistant a question, you will in the best case receive a conclusive and direct answer. Google can also show search results and in the worst case you will hear “Sorry, I don’t know”. Google wants to improve the way we interact with its Assistant, because AI should make conversation much more natural, the company said.

‘Language Model for Dialogue Applications’, or ‘LaMDA’ for short, is the name of the technology that Google is working on. It is about AI in which the neural networks have analyzed a large amount of natural conversations. The intention is that the Assistant not only communicates dry facts, but can also talk similar to a human. In a conversation you start with one topic, the other picks up on it and so you drift off to other things. LaMDA is trained to logically respond to things the user says.

An example of LaMDA

Refine

The Assistant will therefore speak with more nuance and depth, but it is important to Google that the Assistant still continues to say truths. In addition, the company wants to prevent prejudices from creeping into the AI ​​model, taking into account Google’s AI principles. LaMDA is still under development so that the technology can be refined and prepared for use.

During Google I / O 2021, Google also proposed improvements related to the Duplex technology for the Assistant. Google can also automatically replace users’ passwords if they are known to have been leaked. The Android 12 beta has also been introduced with the new Material You design.