Just what the world didn’t need: machine learning applied to advertising

IMAGE: Philipp Schmitt and AT&T Laboratories Cambridge — Better Images of AI (CC BY)

Google has told advertisers in a private presentation entitled “AI-powered ads 2023" that it is working on using generative algorithms to increase the sophistication of personalized advertising campaigns.

The idea is that advertisers will supply creative content that he company’s generative algorithms will be able to recombine based on the information Google has about its users, in the same way — according to them — that an advertising agency could.

I find the idea appalling: on the one hand, it runs the risk of alienating advertising agencies, which will see it as competition when their clients start bypassing them to varying degrees. On the other hand, it poses a new type of personalization by introducing many more variables into the equation, in what is potentially a much more invasive exploitation of user information.

What happens when an algorithm chooses the best way to put an ad in front of you — its layout, its colors, the phrases used, etc. — based on all the information you provide it with, which can include everything Google has on file about you? An algorithm is what it is: the optimization of a given function, in this case, conversion. The danger is not, as Google suggests, that the algorithm will “hallucinate”, but that it will do its job so well that we…