Tech News

Goodbye to “Ok, Google”: now the Assistant will be activated with just a glance

Google announced a new feature to activate your virtual assistant. Is about look and talk which means “look and speak”.


This option will allow the user to interact just by talking and looking at the screen of the Nest Hub Max smart devices. It will not be necessary to say “Ok, Google” every time you want to give an instruction or make an order to the Assistant.

This tool, which was announced during the company’s annual developer event (Google I/O), is now available in the United States for the Nest gadget mentioned above. Once this option is activated, the user will be able to look at the screen and request what they need.


For example, when entering a roomthe user can simply look at their Nest Hub Max and ask it to remind them of a scheduled appointment or to look up some type of data. All this without having to put the famous “Ok, Google” first.

According to Google, the change aims to make conversations with the system more natural.


look and talk it is designed to be activated when the user configures the option on their computer. Both the tools FaceMatch What VoiceMatch, integrated into the Assistant, they will recognize the person’s identity. That way, you know when it’s time to wake up and respond to received commands.

On the other hand, with a view to taking care of security and privacy, the company clarified that the video of these interactions made with the Assistant are processed completely on the device, therefore, the data is not shared with Google or with anyone else.

Quick commands only by voice recognition


Improved the system so that the Assistant is able to understand simple commands or quick phrases such as “turn on the light” or “set the alarm” just by listening to them and, also in this case, without the user having to put the phrase “Ok, Google” in front of it.

In other words, it will be enough for the user to say these sentences, without even looking at the Hub, and the system will carry out the requested action. The user has the option to enable or disable this “quick phrases” feature. If you activate them, the system will be able to respond to requests when it recognizes the user’s voice, thanks to Voice Match.

The Assistant now also recognizes doubts

When speaking, it is common to make pauses or expressions of doubt such as “um” or other interjections that indicate that the user is taking time to finish his sentence, or find the correct word. Between humans this is easy to process and understand but it is not so simple for artificial intelligence systems.

The novelty is that, thanks to the new developments implemented, now the Google Assistant will be able to understand those pauses and expressions of doubt in the speech. During the event, practical demonstrations were made to demonstrate this capacity and the system was capable of interpreting those spaces of silence in an adequate way.

behind the scenes

At the I/O meeting it was explained that developing this function required a breakthrough in artificial intelligence. They explained that six models of machine learning to process more than 100 camera and microphone signalssuch as proximity, head orientation, gaze direction, lip movement, context recognition, and intention classification.

All of this happens in real time so that the device is capable of capturing, recognizing the user and then proceeding to respond to their requests.

Another interesting point is that last year Real Tone was integrated, a tool that allows users to better recognize the different skin tones.

And with the aim of deepening the scope of this option, the Monk tone scale presented at this event will begin to be used, which is an optimization of this function so that the systems are able to identify diversities even better.

More powerful language comprehension models

The possibility that the Assistant is able to understand more subtleties such as pauses in conversations is the result of the development of language processing models increasingly optimized.

We’re getting closer to real-time conversation fluency with the Tensor chip, which is custom designed to handle machine learning tasks on the super fast device.

Little by little, the Assistant is able to better understand the imperfections of human speech, such as interruptions, interjections of doubt and other elements, which makes it possible to have more natural conversations with the artificial intelligence system.


Back to top button