Gmail is Now Blocking 100 Million Extra Spam Messages Every Day With AI (theverge.com)

Google has recruited its in-house machine learning framework, TensorFlow, to help train additional spam filters for Gmail users. With the new filters in place as of last month, the company claims Gmail is now blocking an extra 100 million spam messages every day. From a report: In the context of Gmail's 1 billion-plus users, this isn't necessarily a huge gain -- it works out as one extra blocked spam email per 10 users -- but Google says Gmail already blocks 99.99 percent of spam, so working out what constitutes that last sliver of a percentage is hard.

Comments

Props when deserved

[TFlan91]

Use gmail for work for decade+, spam has certainly never been a problem.

Although, it does seem to be a bit aggressive sometimes.

And the result is more false positives

[kbahey]

And, the result is more false positives ...

In the past 6 to 8 weeks, I found several emails from people I know in the Spam folder.

The strange thing is that those were from email addresses are in my contact list, and have been communicating with me for years.

Bad move Google ...

Re:And the result is more false positives

[N7DR]

It's even worse than that (for me, at least). In the past week, google has bounced several of my perfectly ordinary e-mails to family members with gmail accounts. As far as I can tell, there is no way to tell the AI that it's an idiot. I can't even begin to guess what it thinks is spammy about the e-mails that it's blocked. They don't seem to have anything in common except that they came from me.

The bizarre thing is that (so far) if I re-send the exact same e-mail a few minutes later, it doesn't bounce.

Re:And the result is more false positives

[Solandri]

The strange thing is that those were from email addresses are in my contact list, and have been communicating with me for years.

Being in your contact list is not a white flag. A common tactic of spam malware is to send the contact list of infected person's PC to the spam author. They then spam everyone in the contact list using the owner of the list as the From: email address, precisely in the hopes that an email "from" someone you know is more likely to make it past your spam filter.

Re:Statistical filters... not AI.

[jellomizer]

Because of Buzzword.
When I started working, I help code Decision Support Systems, then they Became Business Intelligence Systems, Now they are called AI.
Of course I am not sure if My job is an Application Architect or Full Stack Developer?
 

Blocked or filtered to Spam folder ?

[ardmhacha]

I have had a GMail account since the invitation only days. I actually got an invite from someone giving them away on slashdot. I am careful about where I use the address and I get very little spam (about 10 / month or less).

When I read about GMail's great spam filters I wonder does that mean that the spam emails are actually blocked (ie. never make it to me at all) or just filtered to my Spam folder.

Re:Blocked or filtered to Spam folder ?

[RhettLivingston]

I've probably been using gmail as long as you. Might have even got the invitation from the same place.

I'm fairly positive that all spam goes to the spam folder. I can't remember the last time I had to mark a spam message. I do look at the spam folder periodically for its entertainment value and to keep up with the flavor of the current scams. I haven't had a false positive in many years.

Re:Statistical filters... not AI.

[Anonymous Coward]

if it's software it's ai
if it's material it's 3d printed

I spend a lot of time in the "Junk" folder now

[Tablizer]

It's making my business negotiations with Nigerian princes very difficult. Maybe I should go into boner pills instead.

Time for more real intelligence?

[shanen]

You managed to remind me of what I still think is the best solution approach. It's actually another aspect of MEPR (Multidimensional Earned Public Reputation). In this application, the dimensions of concern are those that would indicate a particular person with ACTUAL human intelligence is (or is not) skilled at recognizing spam.

One implementation (but the google will never implement it) would be as an opt-in spammer-fighting tool beyond the incredibly naive Spam button. (The additional phishing-report option is not a real solution, but it's a tiny bit better than nothing.) The spammer-fighting tool I fantasize about would parse the suspected spam and let the human being help assess the various aspects of the spam (and recommend appropriate countermeasures) in a kind of automated dialog. MEPR is relevant so that the email server or website can weight the responses and respond appropriately.

Now it's time for some defeatist moron (or shill) to pop in with the idiotic options-based form letter explaining why you can't fight spam. (I'll ignore it, of course. I'm much too stupid to just give up--but again I remind you of the pump-and-dump stock scam spam.) Therefore I'll just bid you ADSAuPR, atAJG.

Re:Gmail is used to spam others though

[shanen]

Hmm... I don't have hard stats but my feeling is that the spam that actually originates within Gmail is a relatively small part of it. However there is a large fraction of spam that uses Gmail accounts as the drop boxes for the suckers, either with Reply-to: headers or in the body of the pitch.

Since I do most of my spam analysis within Gmail, it is not clear if that is a Gmail-specific problem (of better spammer support) or if the spammers just use that approach for each target email system. In that hypothesis, they would favor Gmail drop boxes for spam to Gmail addresses, but favor Outlook drop boxes for spam to Outlook email addresses, and so on.

(I do have accounts on a couple of other email systems, partly as legacies, but mostly to see if any of them EVER get around to offering decent anti-spam features. Most of my Outlook spam is actually fake identity stuff, but I get almost no email of any kind there.)