Artyficial intelligence: what does creative AI mean for marketers?

It wasn’t meant to hap­pen like this. Yes, the robots were always going to come for everyone’s jobs, but it was the menial ones that were set to go first. Freed from the need to fill out spread­sheets and per­form admin­is­tra­tive duties, we were all sup­posed to have extra time to indulge in more cre­ative, ful­fill­ing pur­suits. Yet Microsoft Excel still exists while AI algo­rithms are pro­duc­ing works of art that are both com­mer­cial­ly ­viable and crit­i­cal­ly respect­ed.

An AI artist, Jason Allen, recent­ly caused out­rage among old-school dig­i­tal artists by win­ning a dig­i­tal art com­pe­ti­tion. One of the writ­ers of US pub­li­ca­tion The Atlantic, Char­lie Warzel, pro­voked the ire of illus­tra­tors around the world by choos­ing to adorn an arti­cle about con­tro­ver­sial radio host Alex Jones with an AI-gen­er­at­ed car­i­ca­ture as opposed to using a stock pho­to or com­mis­sion­ing a por­trait.

Three new prod­ucts – OpenAI’s Dall‑E 2, Mid­jour­ney and the tru­ly open-source Sta­ble Dif­fu­sion – have trans­formed the way we see creati­vity since their launch­es this year. Giv­en a sim­ple text prompt, they can cre­ate remark­ably accu­rate visu­als in an infi­nite num­ber of styles. From super­im­pos­ing Ker­mit the Frog on to some of cinema’s best-known ­images to imag­in­ing an ­expand­ed ver­sion of Johannes Vermeer’s Girl with a Pearl Ear­ring, they are ­astound­ing visu­al play­grounds.

Nan­do Cos­ta, direc­tor of Google’s design out­reach team, is opti­mistic about the cre­ative oppor­tu­ni­ties being opened up by these new ­tech­nolo­gies. He argues that tools such as Dall‑E 2 and Mid­jour­ney will “change the way we do every­thing across cre­ative pro­duc­tion”.

But it’s unclear whether any­one tru­ly grasps the sig­nif­i­cance of these advances at such an ear­ly stage. Dis­cussing AI art on their pod­cast, My First Mil­lion, inter­net entre­pre­neurs Sam Parr and Shaan Puri com­pared the arrival of these tech­nolo­gies to pass­ing a pur­ple uni­corn on the road and not stop­ping to make sense of what you have seen. ”Dream” or “gen­er­a­tion” engines, as they called the likes of Dall‑E 2, are indeed some­thing we can recog­nise as spe­cial with­out nec­es­sar­i­ly being able to under­stand their impli­ca­tions.

If the out­put is good enough for both expert eyes and the main­stream media, their appli­ca­tion for com­mer­cial pur­pos­es should be straight­for­ward, right?

For some design, brand and PR agen­cies, the poten­tial is clear. “This is the future of cre­ativ­i­ty,” ­argues Melanie Com­mey, cre­ative ­direc­tor at Bound to Pros­per. “It gives non-design­ers a new oppor­tu­ni­ty to cre­ate and blend con­cepts, offer­ing the capac­i­ty to gen­er­ate ­imagery in mul­ti­ple styles.”

Could this effec­tive­ly reduce time spent on design or reduce the man­pow­er need­ed? “This is depen­dent on how the tech­nol­o­gy is inte­grat­ed in agen­cies over time,” Com­mey says. “But it def­i­nite­ly has the abi­lity to change the nature of cre­ative work as we move to a more vir­tu­al­ly focused realm.”

For oth­er play­ers in the space, AI can serve as a well of inspi­ra­tion – at least, that’s how Dia­gram is ­attempt­ing to use it. The firm’s AI-based sys­tems aid automa­tion via ­design tools such as Fig­ma.

“AI aug­ments and enhances human cre­ativ­i­ty,” says Diagram’s founder and CEO, Jor­dan Singer. “Our AI will extend design­ers’ exist­ing cre­ativ­i­ty and embed itself in their work­flows to allow them to work faster and iter­ate on ideas in tan­dem with the team.”

After all, the impres­sive out­puts from these AI art gen­er­a­tors are pos­si­ble only because human artists and pho­tog­ra­phers orig­i­nal­ly per­fect­ed each of the styles and sub­jects depict­ed. In that sense, these image-gen­er­a­tion toolk­its are use­ful sieves, enabling us to ­explore recom­bined ver­sions of what we’ve already seen. They can become an aid, just as libraries and Wikipedia help writ­ers with their research. But these tech­nolo­gies are clear­ly not with­out con­tro­ver­sy.

“Every tool is per­mit­ted and AI is hap­pen­ing one way or anoth­er,” observes David O’Reilly, an award-­win­ning artist and ­ani­ma­tor. “But this species derives its entire val­ue from the cre­ative work of uncred­it­ed and unwill­ing par­tic­i­pants.”

AI algo­rithms are trained on images and accom­pa­ny­ing texts from across the inter­net. For many peo­ple, this root data belongs to our shared human her­itage. And a lot of it would orig­i­nal­ly be the work of indi­vid­ual artists. Yet, when you use Mid­jour­ney or Sta­ble Dif­fu­sion to illus­trate an arti­cle, there is no cred­it award­ed to the cre­atives who orig­i­nal­ly fuelled the train­ing data. This makes many artists, even those work­ing in the dig­i­tal sphere, wary. There’s a dif­fer­ence between hav­ing your job tak­en away by a ­machine that’s sim­ply bet­ter at mak­ing art and being usurped by one that uses mate­r­i­al stolen from you and count­less oth­er artists.

The speed and rel­a­tive­ly low cost of gen­er­at­ing high-qual­i­ty out­puts is the most imme­di­ate­ly trou­bling aspect for illus­tra­tors and oth­er cre­atives. For edi­tors and mar­keters, it’s a cheap option for bespoke con­tent. For free­lance illus­tra­tors, it might mean indus­tri­al-scale redun­dan­cy in real time.

Nonethe­less, the oppor­tu­ni­ties for stream­lin­ing design briefs, try­ing out new ideas and replac­ing dull stock imagery will make life dis­tinct­ly eas­i­er for many in the pro­fes­sion. But it is unclear whe­ther the world of visu­al cre­ativ­i­ty will nec­es­sar­i­ly be rich­er for this new tech­nol­o­gy in the long term. After all, even the text prompts them­selves can be auto­mat­ed.

“The ques­tion is no longer whe­ther a com­put­er can make art, but whether a human is need­ed at all,” O’Reilly says. “AI is like­ly to lead to unimag­in­ably beau­ti­ful art and pow­er­ful affir­ma­tions of human­i­ty, as well as wide­spread job dis­rup­tion and the atro­phy­ing of fun­da­men­tal human skills.”

The ques­tion remains: will that out­come be worth it?