The fight against spam often resembles the fight against windmills. We seem to be able to limit their number, but unwanted messages will always find their way. Google intends to deal with this problem using advanced machine learning.
TensorFlow probably does not say much about the majority – I do not meet with this name very often. It is artificial intelligence that can be used on many levels. Now the technology created by Google itself will help to fight spam in Gmail.
The artificial intelligence of Google is based on a technique called “deep learning”. And it is this mechanism called by Google TensorFlow that is completely released. It has a neural network (or if someone prefers – an artificial brain) that can collect data and on the basis of it improve its effectiveness and efficiency. Compared to the first generation of Google’s artificial intelligence engine, it is expected to be up to five times faster. Google ensures that it can be run on hundreds of thousands of computers or … a single smartphone.
Every month, 1.5 billion people use Gmail. It is impossible to imagine an e-mail traffic of such a large scale. And all this is spam, which unfortunately still goes to our virtual mailboxes. Google claims it currently offers resistance to spam, phishing emails and malware at 99.9%. The recent implementation of TensorFlow in Gmail services is to help eliminate the 0.1% gap. Already, the new ML (machine learning) helps in capturing an additional 100 million spam messages.
Where are the big numbers from? Until now, Gmail has not done well to filter spam from image-based messages, embedded content or from new domains that hide junk content in normal circulation. The use of TensorFlow in Gmail allows you to eliminate them without fear of accidentally blocking important correspondence. TensorFlow supports current machine learning libraries, and its implementation makes the whole process of learning and filtering new categories of spam easier and more effective. Do you have a problem with spam? write in the comment below.