Last year Qualcomm began with the deployment of the first chips for Android that are characterized by GPU drivers that can be updated to optimize their performance.
Something similar to what happens in PCs where every little bit the drivers or graphics drivers are updated to improve the performance of the new games that are coming out on the market.
It appears that Qualcomm will do something similar for machine learning and AI on the device. During Google I/O 2021, both Qualcomm and the Big G announced a series of upgradeable neural network API drivers.
This new model will be brought to Android 12 and will allow companies or manufacturers to quickly deploy updates through Google Play Services.
Best of all, these updates will be available for older chips alongside older Android versions. In other words, mobile phones from years ago will be able to benefit from these advantages in machine learning and AI.
The NN API (Neural Network API) could improve performance as long as the device had two additional cores, which translates into more efficient use of power while reducing its temperature.
Google Maps and Google Assistant are two applications that use this NN API, and we can understand the benefits that they will bring to both current and old mobiles.
In fact, these two applications will receive significant improvements in search results thanks to the new AI that Google unveiled in I/O and which is known as ‘LaMDA‘. Google Maps will be able to give more ‘smart’ results when we look for the best route with the most beautiful panoramic views. The interesting thing is that part of these processes would occur on the same mobile without having to upload data to Google’s servers.
In short, we will have apps of smaller size in MB that work better and do not consume as much battery for tasks such as voice or face recognition. They would even avoid uploading data to a server by performing all these processes from our same device.