Big Bad Data

“Meet Albert,” the PR pitch read, “the best artificial intelligence friend an advertiser can have.”

Whoa -- talk about anthropomorphism.

“Albert” is Adgorithms’ artificial intelligence-based ad buying platform. The company claims Albert can create, launch and run display, mobile and video campaigns. I wanted to interview Albert, but that wasn’t possible, so I connected with its handler: Adgorithms CEO Or Shani.

Despite the pitch that Albert is omnipotent in the world of ad-buying, Shani said the platform is “designed only to enhance performance, not undermine an advertiser’s decision.” Shani -- like others -- pointed to mankind’s perception, creativeness and ability to strategize as reasons the machines aren’t taking over. “Data-driven machines,” he pointed out, “will have a hard time figuring this out.”

Despite lacking human perception, artificial intelligence machines like Albert offer a different kind of awareness: the ability to analyze vast amounts of digital advertising possibilities, including what to buy, where to buy and who to target. But while AI can interpret Big Data much faster than humans, it’s no less susceptible to fraud. In fact, it may even be more prone.

A large-scale video ad fraud scheme was recently uncovered, and it was designed to deceive machines and humans alike. Programmatic ad buyers were particularly vulnerable to the plot because the fraudsters turned Big Data into Big Bad Data.

The fraud scheme relied on video ad placements that looked like legitimate, high-quality inventory. The illegitimacy was masked by way of sites featuring prominent, above-the-fold video players, video content that targeted high-value verticals and phony banner ads that made the site appear to be a hoppin' place.

In other words, the fraudsters used the ad industry’s love of machines against itself. If your machine-learning, predictive programmatic media-buying Big Data artificial intelligence platform says the inventory is good, who's to argue? (OK, so not all advertisers are blindly trusting technology, but they need to be careful not to drown in jargon-laden Kool-Aid.)

We can’t just get rid of machine-based media trading, because despite the new issues that have surfaced, the technology has buried many old problems. In truth, it’s quite the catch 22. Machines are being used to trick other machines, and the only thing that can effectively catch this skullduggery is -- you guessed it -- machines.

"Fraud" image from Shutterstock.

Recommend (3) Print RSS
1 comment about "Big Bad Data".
  1. Paula Lynn from Who Else Unlimited , July 2, 2014 at 3:46 p.m.
    HAL. HAL. HAL. BOOM