
Google is experimenting with a big change in how machine-learning systems work on Android.
Google is testing a new way to train its artificial intelligence algorithms on users Android phones. The search giant is using data collected from its keyboard app, Gboard, to modify users’ search results. But firstly, lets understand the basics of what Google actually does.
No one will deny the fact that Google knows everything about you. The search giant is not only giving free information from all over the world, but at the same time it is also collecting a large number of data from people using it. The data is collected in one place and changes are made to the algorithm according to the data collected. Hence the final result where you can search better and the information comes more handy and accessible to users. While this method is impressive in many ways, the only hurdle Google has to deal with is that the data and feedback has to be saved on their server which might cause security concerns. So, in order to not risk users personal information, Google is experimenting with a new method of AI training it calls Federated Learning.
As mentioned earlier, Google is using its Gboard for this training purpose. When Gboard shows users suggested Google searches based on their messages, the app will take note of what suggestions users is taking and at the same time it will also notice suggestion users ignored. That data will then be used to personalize the app’s search results on each individual’s device. Which means users personal information will be secure as the new algorithm will be trained in their devices only.
You won’t have to update your device to get the latest update. Google will collect all data and incorporate them into a single new update to the app for all users.Your algorithm will begin rolling out on the app immediately after data is collected. On one hand where Google collects a lot of personal information, Google says it’s actually more secure to do it this way as the data used to improvise your search and browsing result will never leave your phone.
Google says the whole system has been streamlined to make sure it doesn’t interfere with your phone’s battery life or performance. The training process only takes place when your phone is “idle, plugged in, and on a free wireless connection.” The Google blog post further mentions that to departure such a network to a huge audience, the company uses on device training version of TensorFlow.
Well Google is not the first tech giant to invest in AI learning. Apple, last year invested around $200 million to buy a machine-learning startup called Turi.