Media Buying Briefing: Artificial intelligence ‘is gonna ruin the world… and then we adapt’

By Michael Bürgi

Marketers and agencies love to throw around “artificial intelligence” and “machine learning” when talking about raising the bar of effectiveness to new heights, but too often that talk feels like more hype than actuality. That may be starting to change, as AI and its variants find their place in the buy-sell-market equation to run multiple tasks in a blink, unearth insights in troves of data, and even get involved in creative and the placement of advertising.

Starting with the creative process, AI has the potential to get used in the creation of “derivatives” of an original creative concept, explained Rob Tan, CTO of native programmatic firm Sharethrough. As he told it, a human creative will for the foreseeable future craft an original concept for content or advertising, but AI can quickly and effectively make innumerable spinoffs of that concept.

“AI will pull away some of the reliance on creative agencies,” said Tan. Once an agency creative department crafts an original concept, “they’re used for all these minor tweaks in copy, images and other things. [Instead], AI can create very meaningful derivatives, alterations that make perfect sense. And may in fact take into account other factors to ensure it’s a compelling headline or image that’s automatically selected.”

Tan pointed to AI-driven technology such as GPT-3 (generative pre-trained transformer 3, which is an autoregressive language model that uses deep learning to produce human-like text) as key to such advancements. With GPT-3, “you can get it to respond the way you want it to. It’s almost creepy how you can have AI talk to you like a human does,” he added.

AI tech has found its most comfortable and useful place in media on the data and analytics side. (Tan actually described what he dubbed “the media industrial complex” as generally resistant to trying out AI solutions in buying and planning.) Crossmedia’s managing partner Lee Beale said the agency’s data unit, Redbox, makes use of machine learning models for attribution measurement and predictive modeling. These techniques are applied to large data sets like log level data from universal ID graphs, as well as clients’ first-party data for modeling lifetime value multipliers or churn risk.

“These models run daily and, like AI promises, get smarter (more predictive) with increasing observations of events,” said Beale in an email.

Redbox also delves into optimizing creative use and execution as well “to tease apart the elements [that] drive results,” added Beale. “This involves the use of multiple AI platforms which discern physical and emotional components of static and video creative, breaking them down into a myriad of component part probability scores, which are then modeled against business outcomes.”

Third-party AI firms also help agencies uncover new insights for their clients. AI firm Helixa offers a SaaS product (used by a number of agencies including VaynerMedia, Rauxa and VMLY&R, as well as marketers and media companies it declined to identify) that sits between traditional research companies and social listening tools, but identifies actual behavior rather than through panels and surveys, to find complex patterns …read more

Source:: Digiday