Skip to main content

Google's AI is warming up for world domination by killing inbox spam

Google's AI is warming up for world domination by killing inbox spam

Share this story

News about artificial intelligence currently tends to fall into two main camps. In the first, we have technologists and computer experts getting genuinely worried about the future threat of out-of-control AI. And in the second, we have the novel but ultimately modest uses the technology is actually being put to in the present day. Google's announcement this week that it's using machine learning technology to keep your Gmail inbox free of spam definitely belongs in the latter category.

In a blog post published yesterday, Google product manager Sri Harsha Somanchi explains that Gmail's spam filter now uses "an artificial neural network to detect and block the especially sneaky spam — the kind that could actually pass for wanted mail." The company is also deploying its machine learning tech to adapt to individuals' inbox choices, essentially creating custom spam filters on-the-fly. "So while your neighbor may love weekly email newsletters, you may loathe them," writes Somanchi. "With advances in machine learning, the spam filter can now reflect these individual preferences."

less than 0.1 percent of email arriving in the average gmail inbox is spam, says google

It may be difficult for Google to explain exactly how its neural networks filter out unwanted mail, but at least the results are clear. The company boasts that less than 0.1 percent of the email arriving in the average Gmail inbox is now spam, and that the amount of wanted mail falsely identified as spam is under 0.05 percent. This seems like a relatively small improvement from 2012 (when the company claimed that less than 1 percent of incoming mail was unwanted), although even a single wrongly-marked email can still be frustrating to users.

AI, it turns out, is pretty good tool at this sort of filtering. Facebook has been researching how to use artificial intelligence to trim your News Feed since 2013, and according to a report from Wired this week, Twitter's engineers are also exploring how to use AI to filter out unwanted messages. This is all well and good for keeping today's internet users happy, but it also provides us with a neat early-warning system for future AI takeovers. You just know that the last thing the internet will experience before Skynet turns itself on will be a torrent of penis enlargement and scam emails as our ambitious computers focus their attention on world domination. At least that'll keep us busy during the apocalypse.