In recent months, Google infused several of its products with machine learning and AI applications. From Adaptive Battery in Android Pie to the recent Shadow Art web app, the search giant uses AI a lot.
Most recently, Google added machine learning to its robust spam filters in Gmail. According to the Mountain View-based company, the new protections catch and an additional 100 million spam messages every day.
When you consider the scale of Gmail, that’s not exactly a huge number. With 1.5 billion monthly users, that works out to about one message blocked for every 15 users.
However, Google claims it already blocks more than 99.9 percent of spam, phishing and malware emails, so finding an extra 100 million spam messages is no small feat.
The company used its open-source machine learning framework, TensorFlow, to block spam categories that are typically difficult to detect. This includes image-based emails, emails with hidden embedded content, and messages from newly created domains that try to hide a low volume of spammy messages inside legitimate traffic.
Further, TensorFlow allows Gmail to adapt its spam filtering to individual users, as not everyone considers the same emails to be spam.
So, if you’ve noticed fewer spam emails recently, thank TensorFlow.