Can a robot build your creative? Maybe …

August 28, 2010 by Dave Haynes

A piece in the Media and Advertising section of the New York Times this morning discusses a provocative proof of concept automated advertising solution put together by the Paris-based creative agency BETC Euro RSCG, which is part of the Havas agency empire.

The software is called CAI, pronounced Kay, for Creative Artificial Intelligence, reports the Times.

CAI can be programmed to produce ads by selecting a product category (say, soft drinks) and type of product (for instance, coffee, energy drinks, fruit juice, milk, tea or water).

Next up are questions about objectives. Do you want to generate awareness? Create loyalty? Increase purchase? Introduce a product? Recruit customers? CAI then wants to know the demographic target for the ad by sex and age.

Last come questions on the intended benefits of the product. For milk, for example, qualities like fresh, healthy and organic are offered. CAI ponders all those requirements, then produces three possible ads that meets them.

CAI can randomly generate an estimated 200,000 ads. In a recent demonstration, the software brought forth bland and formulaic — but perfectly acceptable — ads that could run in magazines or newspapers, as banners on Web sites or on billboards.

The guy behind this effort – Stéphane Xiberras, the President and Executive Creative Director, BETC Euro RSCG/Paris – has written that the project came about to prove a point – that most advertising work that gets pushed out each day is uninspired rehashes of earlier work. Most of what we see all day, he theorized, could be done by a robot.

What started out as an intellectual game became more and more alarming as we progressed in the development of our robot. We expected to create a clumsy rather grotesque machine that would be systematically way off the mark, in a comical way.

Unfortunately, this is not the case. CAI produces something that is caricature, but that very often by some random diabolical grace, reminds us of an ad we have already seen on a street corner or on the page of a magazine.

CAI creates something that resembles advertising, but that fundamentally isn’t, in the sense that it lacks essential qualities: novelty, inventiveness and the unexpected. In short, anything that only a human being is capable of producing. CAI is a fascinating but dangerous machine because it synthesises the nemesis of our creative profession: standardised or formatted thinking, call it what you want.

In this sense, it is our responsibility to cultivate the differences between our agencies, to encourage our planners and our creatives to always strive to go further, and never to settle for regurgitating what has already been seen.

I definitely agree that most of the advertising work that we watch, see, hear and read day to day represents variations on very common, time-honored approaches. It’s not particularly inspiring or impactful. Point taken and somewhat proven by CAI (Kay).

Xiberras suggests this should remind agencies and their clients how essential “creative” is in advertising production, and that should put an end to any notion that advertising can be automated.

I don’t agree there.

1 – The agency business is not set up to develop and deliver high-concept, high production value, high cost work. It is not set up to do what is the majority of advertising out there – which tends to get produced by very small shops or in-house by the media properties. Newspapers, my longtime frame of reference, have creative departments that build a lot of the stuff that fills pages.  There is a big gap between what agenecies want and need to charge and what local advertisers are ever going to pay.

2 – The CAI model appears to do print, but there is already stuff on the market that largely automates video generation (like Animoto). For a pretty healthy chunk of the small to medium business market that does advertising, they’re not looking for or getting much more than “Weekend Sale!!! — 50% off!!! — Come On Down!!!” with a lot of pops, zooms and fades in a 15-30 second spot. I’d at least suggest that could indeed by automated. The stuff will never win awards, but can do the basic job of making shoppers and customers aware of an offer or service.

Stuff like FlyPaper and Animoto get part of the way there. Someone, somewhere, is going to come out with a version of CAI that is not a theoretical model lament the dearth of innovation, but a commercial platform that is spitting out uninspiring but workable spots for $XX a pop.