
AI-based spam filters are better than traditional ones because they improve as they go along.
Google recently announced that advances made in the neural networking field and artificial intelligence helped its e-mail service identify and lower Gmail spam rate to 0.1 percent from the 1 percent level reported by the company three years ago.
Artificial intelligence also helped the company’s e-mail service to identify only 0.5 percent of friendly mails as spam mails, from an one percent level reported in 2012.
Google said that the current rates are the best current technology can provide and they are mainly due to human brain-like neural networks-based technology which analyzes and compares heaps of data gathered from an enormous database of computers to tell genuine e-mail messages from spam and phishing mails.
“One of the great things about machine learning is that it adapts to changing situations,”
noted John Rae-Grant, a senior product developer for Gmail, which according to the company currently nears 1 billion accounts.
Mr. Rae-Grant said that the current technology used to filter spam does not rely anymore on spam filters that use stiff rules. Instead, Google employs AI which generates new rules as the spam filter encounters new challenges day by day.
The company also used neural networking technology for voice recognition purposes with Android, face recognition with Google Photos and so on. Facebook and Twitter are also digging deeper into the technology. Facebook announced last mouth that it would use it to improve the News Feed.
Moreover, the alternative to Google’s search engine in China – Baidu– uses artificial intelligence tools to learn what types of ads users might be interested in. Yet, when detecting e-mails, the algorithms try to figure out what you might NOT be interested in.
Twitter announced that it would use AI to automatically detect spam messages and trolls.
But bringing junk mail near to zero was a decade long effort. Three years ago, Microsoft managed to remove almost all spam from Hotmail inbox. Only 3 percent escaped the service’ spam filters.
At that time, Google boasted that its spam filters could reduce spam to 1 percent, while the rate of false positive messages was also 1 percent. False positive means that a real e-mail message is mistakenly flagged as junk mail by spam filters.
Both companies used in 2012 heuristic algorithms that could spot spam on preexistent rules. Yet, although the technology was a huge step ahead, it was still not enough. The false-positive rate was especially high because you had one chance in 100 e-mails to miss a legitimate message, which could have been crucial.
Image Source: Digital Trends