Not this kind of spam. Credit: Matthew W. Jackson.

Kaspersky Lab reported last week that spam email is on the decline, having fallen from 80%-85% of all email over the past few years to about 65% at the end of 2012. There are two main reasons for this. First most spam filters are at least 98% accurate. That may not be so great for a human, but it’s astoundingly good from a computer programming perspective. Second, most email clients are demanding properly-encrypted DKIM signatures, which verify the email address of the sender

Together, these two things make it almost impossible to get spam to your inbox, and it is becoming unprofitable. Many spammers have taken note and moved on to other venues, like legal internet ads, coupon services, and group discount websites. Caveat emptor.

Some of the more clever spammer (or scammer) methods are still in play, like malicious Google Documents and fake personal correspondence, but from the sudden drop in volume, it looks like we may be headed for a world (and may be reaching it, from the end users’ perspective) where no one really worries about spam anymore.

At least, that’s the good ending.

But there’s another side to this. There will always be scammers out there, and their computers are getting smarter just as fast as ours. Suppose that instead of trying to sneak around the spam filters by throwing in lots of random words or words with ch4racter substituti0ns, which appear far less legitimate to human eyes, artificial intelligence becomes advanced enough to write bogus emails that look like they were written by humans. Then, not only could the filters have a hard time spotting them, so could we.

To be sure, things will be harder at the senders’ end. Address verification means it could take legions of gmail address or even buying a few domain names to send large amounts of spam, and people will always be wary of unsolicited messages from someone they don’t know. Another proposed reform is the proof-of-work system, which slows the computer down for a few seconds when sending a message. This would be almost invisible to a normal user, but could slow down a spam-bot by orders of magnitude. It’s not clear who will have the upper hand long-term.

So what do you think? Will we stay ahead of the curve on this, or will spam make an AI-powered comeback in the future?