Google’s Gmail is able to accurately decipher which of your emails are of primary importance, which are necessary updates, which are commercial promotions, which are social media notifications, and which are downright spam. Their technology already blocks 99.9% of spam by employing both AI and rule-based filters, according to the Verge, but TensorFlow, the machine-learning AI launched by Google in 2015, placed new Gmail filters last month that are blocking an extra 100 million spam messages every day.
Considering Gmail has over one billion users, 100 million isn’t that consequential to an individual user. It adds up to only be one extra blocked spam email per 10 users on the Google platform, which will likely go unnoticed by almost all Gmailers. The advance, however, does have exciting implications for the future of Google’s AI capabilities.
“At the scale we’re operating at, an additional 100 million is not easy to come by,” Neil Kumaran, product manager of Counter Abuse Technology at Google, tells The Verge. “Getting the last bit of incremental spam is increasingly hard, [but] TensorFlow has been great for closing that gap.
As a free AI that allows developers to apply it to a variety of tasks, TensorFlow is known for its flexibility, capacity to scale, and compatibility with other Google programs. Such comprehensive technology has significantly boosted Google’s place in the AI market by encouraging users to buy computing power, off-the-shelf vision, and speech algorithms from Google.
TensorFlow’s flexibility also makes spam filters more customizable so that users can rely on AI to keep out spam but can also have a say in what exactly spam is to them. “There’s no one definition of spam out there,” Kumaran continued, highlighting the importance of user involvement in spam filtering to help the program in “turning those signals into better results.”