Eliminating The 'Learning' From 'Machine Learning'

Why learn, when you can be omniscient from the start?

That’s what Tribal Fusion is asking, as it launched a new technology this week meant to speed up the “learning” bit of “machine learning.”

Actually, the company wants to accelerate the learning curve so much that you never even notice it. 

Tribal Fusion, a division of Exponential, has christened its new technology AERO. (Really, we should be glad they went the acronym route here, as what it stands for is hyperbabble-worthy: Audience-Efficient Real-time Optimization.)

It’s pretty thick stuff, but here’s the gist of it: AERO is an algorithm meant for display advertising campaigns. Tribal Fusion says the algorithm applies Exponential-owned behavioral audience models to campaigns from the start, meaning advertisers don’t need to wait for their campaign to run for a little bit while its algorithms figure out the best audience.

In order to do this, a company needs an existing bank of data and audience models -- as Tribal Fusion has by way of Exponential. I would argue that building those data sets and creating those audience models count as a learning process, meaning the “omniscient from the start” part is only a half-truth. Plus, doesn’t every advertiser have first- or third-party data they apply to campaigns before launch? No advertiser runs a campaign blind from the beginning.

Regardless of whether this really is a brand new type of technology -- one that helps advertisers optimize a campaign from day one -- this is the first time I’ve seen a company try to change “machine learning” into “machine.”

"Algorithm" image from Shutterstock.
Recommend (9) Print RSS
1 comment about "Eliminating The 'Learning' From 'Machine Learning'".
  1. Pete Austin from Triggered Messaging , April 10, 2014 at 5:04 a.m.
    Re: "No advertiser runs a campaign blind from the beginning". Actually, personal 1-to-1 campaigns are run blind from the beginning, in the quite common situation that you've not seen a target shopper before. For example personalized product suggestions on web pages and in emails. The recommendation engine makes judgements based on past data, but it takes a few page views before these are much more than guesses.